CN121540727A

CN121540727A - A Defect Detection Method for Injection Molded Parts Based on Dual Robotic Arm Collaborative Operation

Info

Publication number: CN121540727A
Application number: CN202610062616.4A
Authority: CN
Inventors: 宋涛; 李宗航; 曹松晓; 徐志鹏; 蒋庆
Original assignee: China Jiliang University
Current assignee: China Jiliang University
Priority date: 2026-01-19
Filing date: 2026-01-19
Publication date: 2026-02-17

Abstract

The invention relates to the defect detection field, and discloses an injection molding defect detection method based on the cooperative operation of double mechanical arms, which has the technical scheme that the synchronous acquisition of double-sided images is completed by the cooperation of an upper mechanical arm and a lower mechanical arm, a turnover link is omitted, the problem of detection efficiency low, product damage and positioning deviation because of the upset leads to among the prior art has effectively been solved, has and need not to overturn the two-sided synchronous detection of injection molding can be realized to the operation, effectively avoids detection cycle extension, product secondary damage and positioning deviation problem, promotes simultaneously and detects the advantage of integrality and production efficiency.

Description

Injection molding defect detection method based on double mechanical arms collaborative operation

Technical Field

The invention relates to the field of defect detection, in particular to an injection molding defect detection method based on cooperative operation of double mechanical arms.

Background

The injection molding piece of the front end frame of the automobile is a key part in the structure of the automobile body, is usually formed by injection molding engineering plastics, has the characteristics of light weight, high strength, corrosion resistance and the like, can integrate various functional structures and reduces the assembly complexity. The main application of the device comprises the steps of supporting and positioning parts such as a radiator, a car lamp and a bumper, ensuring the assembly precision of the whole car, absorbing energy in collision to improve the safety and optimizing the aerodynamic performance. Because the front end frame directly influences the safety, assembly quality and appearance matching of the automobile, defects such as material shortage, breakage, sink marks and the like in the injection molding process of the front end frame must be strictly detected so as to ensure the product quality and the production efficiency. At present, surface defect detection is generally carried out by adopting an image detection mode, surface image information of injection molding parts is shot by an industrial camera and is uploaded to a defect detection system, and the defect detection system is used for judging the surface defects of the injection molding parts. In order to meet the requirement of double-sided detection, a conveying mechanism and a turnover mechanism are generally arranged in a matched manner, the conveying mechanism is used for conveying injection molding parts to sequentially pass through an image acquisition area, after single-sided detection of injection molding parts is finished in batches, the turnover mechanism is used for or manually placing the injection molding parts in a turnover manner, image acquisition is repeated, and then double-sided detection is finished.

However, the manner of turning over the injection molding to achieve double-sided inspection has significant drawbacks. Firstly, repeated conveying and turning operation are needed, so that the detection efficiency is greatly reduced, the high-speed beat requirement of a modern production line is difficult to meet, when the turning treatment of batch injection molding pieces is carried out, partial injection molding pieces are easy to exist, the situation that the detection result of the injection molding pieces is affected because the injection molding pieces are not turned over is easy to cause secondary damage or position deviation of products in the turning process, the detection precision is affected, furthermore, the equipment occupation area is large due to the detection of the twice conveying, the layout of the production line is complex, the production cost is increased, and finally, the requirement on the positioning precision of a conveying belt is extremely high, and any tiny position deviation can cause that the front and back two shots cannot accurately correspond to the same detection area, so that the reliability of the detection result is seriously affected. These drawbacks make the detection method significantly deficient in terms of efficiency, accuracy and practicality.

Disclosure of Invention

Aiming at the defects of the prior art, the invention aims to provide the defect detection method for the injection molding piece based on the cooperative operation of the double mechanical arms, which has the advantages of realizing double-sided synchronous detection of the injection molding piece without turning operation, effectively avoiding the problems of prolonged detection period, secondary damage of products and positioning deviation, and improving the detection integrity and the production efficiency.

In order to achieve the above purpose, the present invention provides the following technical solutions:

an injection molding defect detection method based on double mechanical arms collaborative operation comprises the following steps of

A feeding conveying step, namely conveying the injection part to be detected to a feeding station by a feeding conveying device, grabbing the injection part to be detected by a feeding clamping claw, and arranging the injection part to be detected on an equipment conveying belt, and conveying the injection part to the detection station by the equipment conveying belt;

an image acquisition step, wherein the upper mechanical arm and the lower mechanical arm respectively acquire image information of preset point positions according to a preset acquisition sequence;

The detection step, according to the detection type of the detection point in the image information and adopting a corresponding detection scheme to obtain a detection result, if the detection result is unqualified, storing the detection image and the corresponding detection point into a defect database, and outputting an unqualified signal to an upper computer;

And in the blanking conveying step, the equipment conveying belt conveys the injection molding part to a blanking station, the upper computer judges the quality of the injection molding part according to the detection results of all detection points, if the injection molding part has defects, the injection molding part flows out along with the equipment conveying belt and is defined as an unqualified product, and if all the detection points are qualified, the injection molding part is grabbed by the blanking clamping jaw and is placed in a qualified product area.

In the present invention, preferably, the detection scheme is used for identifying fracture, material shortage, incomplete penetration, deformation and other appearance defects, and includes a slot detection policy, a buckle shrapnel detection policy and a beam detection policy, where the slot detection policy performs a brightness detection policy for a slot or a through hole area, the buckle shrapnel detection policy performs a template matching and edge analysis policy for a buckle or shrapnel structure, and the beam detection policy performs a connectivity analysis policy for a beam area. The scheme solves the problem of insufficient adaptability of the general scheme in complex structure defect identification by customizing different detection strategies for different structural features of the injection molding part. The method comprises the steps of executing a brightness detection strategy aiming at a slot hole area, dividing a detection area according to geometric form characteristics of a slot or a through hole, utilizing sensitivity of brightness change to identify a non-through defect, avoiding interference of material reflection or background noise on hole penetrability judgment, ensuring that normal and abnormal states can still be distinguished stably under the condition of illumination fluctuation, executing a template matching and edge analysis strategy aiming at a buckle shrapnel structure, carrying out local area division according to fine structural characteristics of the buckle or the shrapnel, quickly positioning a standard shape through template matching and combining with edge strength analysis, effectively capturing tiny fracture or deformation, overcoming the defect that the traditional single algorithm is insensitive to edge continuity change, executing a connectivity analysis strategy aiming at a cross beam area, designing a continuity verification mechanism based on linear extension characteristics of the cross beam, accurately identifying fracture or material loss through dynamic evaluation of area connectivity state, and solving the problem of integral functional failure caused by local interruption of a long-strip structure. The cooperative application of the strategies enables the detection process to closely fit the actual structural characteristics of the injection molding part, and the pertinence and the robustness of defect identification are remarkably improved.

In the present invention, preferably, the slot detection strategy includes:

Determining the central position and boundary size of a clamping groove or a through hole in the image information according to a preset structure template, and forming a corresponding slotted hole detection area;

Dividing the slot detection area into a plurality of layers in a direction perpendicular to the opening direction, and further dividing each layer into a plurality of local areas;

calculating the average brightness of each local area in each layer, and taking the average brightness as a layer brightness value;

comparing the layer brightness value with a corresponding normal brightness range, and defining a layer with the layer brightness value exceeding the normal brightness range as a brightness abnormal layer;

Carrying out reflection interference identification and brightness compensation on the local position of the brightness abnormal layer, and then carrying out brightness re-judgment;

When the brightness of the bottommost layer is abnormal and the boundary is discontinuous, the defect that the defect is not completely penetrated is judged.

According to the technical scheme, through a systematic layering detection and interference compensation mechanism, the accuracy of defect identification of the slotted hole area is effectively improved, and particularly reliable judgment of incomplete through defects is achieved. Specifically, the center position and boundary size of the clamping groove or the through hole are determined in the image information according to the preset structural template, and a corresponding slotted hole detection area is formed, so that the detection area is ensured to be strictly limited in the range of the target structure, the interference of background noise or part positioning deviation on subsequent analysis is avoided, and a space reference is laid for fine detection. The slot detection area is divided into a plurality of layers according to the direction perpendicular to the opening, each layer is further divided into a plurality of local areas, the depth characteristics of the holes are closely attached to each other in the dividing mode, the brightness change gradient along the depth direction of the holes can be accurately captured by layering perpendicular to the opening direction, the defect positioning of the microscopic level is realized by subdivision of the local areas, and the integral brightness fluctuation and the local abnormality are effectively separated. The average brightness of each local area in each layer is calculated, the average brightness is used as a layer brightness value, the influence of random noise and tiny reflection points is restrained through local average processing, a more representative layer brightness index is generated, and a stable basis is provided for abnormality judgment. The layer brightness value is compared with the corresponding normal brightness range, and the layer with the layer brightness value exceeding the normal brightness range is defined as the brightness abnormal layer, so that the efficient preliminary screening is realized, the potential problem area is focused rapidly, and the redundant calculation of global scanning is avoided. The method comprises the steps of carrying out reflection interference identification and brightness compensation on the local position of a brightness abnormal layer, and carrying out brightness re-judgment, wherein the step is specially aimed at the common material reflection problem of an injection molding part, and the brightness distortion caused by reflection is eliminated by identifying the reflection mode and carrying out compensation, so that the re-judgment mechanism ensures the rigor of abnormal judgment and prevents the reflection area from being wrongly judged as a defect. When the brightness of the bottommost layer is abnormal and the boundary is discontinuous, the defect which is not completely penetrated is judged, and the judgment logic combines the position sensitivity and the boundary continuity analysis, because the defect which is not penetrated is usually characterized by double characteristics of bottom brightness abnormality and boundary fracture, and other abnormalities such as surface reflection often do not meet the condition, so that the defect is accurately identified.

In the present invention, preferably, the clip spring detection strategy includes:

reading a standard template of the buckle or the spring plate, and determining a buckle spring plate detection area in the image information;

dividing the buckle elastic sheet detection area into a plurality of parts according to structural characteristics, and comparing templates of the parts to obtain similarity;

Performing edge extraction processing on the part with low similarity, and dividing the part into a plurality of local areas;

calculating the edge strength and trend consistency of each local area so as to judge the edge reliability of each local area;

When the edge reliability is insufficient, calculating the offset between the edge position of the area and the edge position of the template;

And if the offset exceeds the preset range, judging that the device is broken or deformed.

According to the technical scheme, the problem of judging reliability caused by position deviation and interference in the detection of the buckle elastic sheet is effectively solved through a staged fine processing mechanism. The method comprises the steps of firstly reading a standard template, determining a detection area, establishing a reference for subsequent analysis, avoiding overall matching failure caused by initial positioning deviation, then dividing the detection area according to structural characteristics, comparing the templates, obtaining similarity of each part, accurately identifying a structural abnormal area in a localized dividing mode, preventing global matching from covering local defects, extracting edges of parts with low similarity and further dividing the parts into local areas, carrying out deep analysis on the areas with focus problems, avoiding redundant processing on the areas with high matching, obviously improving detection efficiency, then calculating edge strength and trend consistency of each local area to judge edge reliability, combining strength and direction double indexes, effectively filtering interference factors such as reflection, noise and the like, ensuring objectivity of edge characteristic evaluation, calculating offset of actual edge positions and template positions when the edge reliability is insufficient, providing quantifiable basis for defect judgment by replacing subjective experience judgment according to difference of quantized positions, and finally judging fracture or deformation defects according to whether the offset exceeds a range or not, and ensuring consistency and accuracy of a thresholding judging mechanism. The method comprises the steps of carrying out edge extraction according to a part with low similarity, concentrating analysis resources in a potential defect area, avoiding the problem of low efficiency of unified processing of the whole area in the traditional method, calculating edge strength and trend consistency, enhancing edge quality evaluation capability through multi-dimensional feature fusion, reducing misjudgment risk caused by illumination change, establishing objective standards of defect judgment by accurate calculation of offset and threshold judgment, and solving the problem of insufficient reliability caused by edge blurring or interference in the prior art.

In the present invention, preferably, the beam detection strategy includes:

Determining the overall position of the cross beam in the image information through boundary extraction and continuity analysis;

dividing a plurality of continuous detection sections along the length direction of the cross beam, and setting a preset overlap amount within a certain range between adjacent detection sections;

performing multi-level bright-dark separation processing on the inside of each detection section to obtain a foreground region;

analyzing the area size and the aspect ratio of the foreground region to form detection characteristics;

sequentially executing foreground region quantity comparison, area occupation ratio judgment and whether foreground regions in adjacent positions are continuous or not;

if a plurality of continuous abnormal sections appear along the length direction, the cross beam is judged to have fracture or material missing.

According to the scheme, robustness and accuracy of beam defect detection are remarkably improved through a layered progressive continuity analysis mechanism. Firstly, the overall position of the cross beam in the image information is determined through boundary extraction and continuity analysis, so that the accurate initialization of a detection area is ensured, the follow-up analysis deviation from an actual structure caused by positioning deviation is avoided, and a reliable foundation is laid for follow-up segmentation processing. Then, dividing a plurality of continuous detection sections along the length direction of the cross beam, and setting a certain range of preset overlap between adjacent detection sections, wherein the segmentation strategy decomposes the long structure into manageable local units, and the design of the preset overlap effectively eliminates the potential defect omission risk at the boundary of the detection sections, so that the continuity analysis can be connected across the boundary between the sections in a seamless manner, thereby overcoming the problem of continuity misjudgment caused by edge blurring in single long-area detection. Then, multi-level bright-dark separation processing is carried out on the inside of each detection section so as to obtain a foreground region, a multi-level processing mechanism dynamically adjusts separation threshold values according to local brightness gradient of an image, the multi-level processing mechanism adapts to complex scenes of uneven reflection and illumination change on the surface of an injection molding piece, defect details lost in a strong reflection region or noise interference generated in a dark region in single threshold value processing is avoided, and the foreground region is ensured to accurately reflect real structural features. And then, analyzing the area size and the aspect ratio of the foreground region to form detection features, quantifying the abnormal degree of the structural morphology by the features, wherein the area size is used for identifying the size of the material loss, and the aspect ratio captures the geometric distortion caused by deformation to jointly construct an objective basis for distinguishing the normal structural variation and the defect. On the basis, the foreground region quantity comparison, the area ratio judgment and the continuous check of the foreground region at the adjacent position are sequentially executed, the random noise interference is eliminated through quantity comparison in the multi-stage check flow, the significance of the defect is checked through the area ratio judgment, and finally the space relevance is checked through the continuity check. Finally, if a plurality of continuous abnormal sections appear along the length direction, the cross beam is judged to have fracture or material deficiency, and the judgment logic requires the abnormal sections to form a continuous sequence in space based on a continuity principle, so that the interference of isolated noise points is eliminated, the reliable identification of the fracture or material deficiency defect is ensured, and the problem that single-point abnormality is easily misjudged as the defect in the traditional method is solved.

In the present invention, preferably, the image acquisition step is completed by the upper mechanical arm and the lower mechanical arm synchronously, wherein the upper mechanical arm and the lower mechanical arm respectively move to the shooting positions according to the preset motion track, and the shooting gesture and the distance are ensured to be stable through the position feedback signals. According to the technical scheme, the image acquisition process is optimized through a double mechanical arm cooperative mechanism and real-time feedback control, and the problems of efficiency bottleneck and image quality caused by unstable posture due to traditional sequential acquisition are solved. The image acquisition step is synchronously completed by the upper mechanical arm and the lower mechanical arm, so that images on the upper surface and the lower surface of the injection molding piece can be acquired simultaneously, waiting time generated by waiting for single-sided detection completion is avoided, the detection period is greatly shortened, the consistency of double-sided images in the time dimension is ensured, and a synchronous data basis is provided for subsequent defect comparison analysis. The upper mechanical arm and the lower mechanical arm respectively move to shooting positions according to the preset movement track, the mechanical arm is ensured to be positioned to the optimal shooting angle quickly and collision-free through a pre-planned accurate path, the uncertainty caused by on-site dynamic adjustment is avoided, and reliable hardware execution guarantee is provided for high-quality image acquisition. The shooting gesture and the distance are ensured to be stable through the position feedback signal, the vibration or the external interference of a mechanical system is dynamically compensated by utilizing real-time monitoring data, so that the relative position between a camera and an injection molding piece is always kept constant, the phenomena of image blurring and deviation are effectively inhibited, the definition and the analyzability of an image are obviously improved, and a solid foundation is laid for the accuracy of defect detection.

In the present invention, preferably, a preparation step is further provided before the feeding and conveying step, including obtaining a model of an injection part to be detected, and the detecting device adjusts a distance between the device conveyor belts, a stop position and the preset point according to the model.

In the present invention, preferably, the feeding and conveying step includes that when a sensor on the feeding and conveying device detects that an injection molding part arrives, the feeding gripper grips the injection molding part to be detected and is placed on the equipment conveying belt. According to the scheme, the position state of the part is monitored in real time through the integrated sensor, a dynamic trigger mechanism based on a feedback signal is constructed, the problem that the grabbing time is inaccurate in the feeding process is effectively solved, and therefore the continuity and the detection precision of the production line are improved. The detection mechanism accurately captures the actual in-place moment of the part by using a physical signal when a sensor on the feeding conveying device detects that the injection molding part arrives, avoids rough operation depending on fixed time intervals or manual judgment, ensures that the grabbing action is started only when the part is stably at a target position, triggers the feeding clamping claw to grab the injection molding part to be detected based on a real-time signal fed back by the sensor, enables grabbing behavior to be closely related to the physical state of the part, remarkably reduces grabbing failure risk caused by position deviation, optimizes production line beats, avoids invalid waiting of the part on the station, and then accurately places the grabbed part on a device conveying belt, ensures that the part is reliably conveyed to the detection station through cooperative cooperation of the sensor signal and the action of the clamping claw, and provides a stable input basis for subsequent image acquisition and defect analysis. The core of the whole mechanism is that the sensor detection is used as a decision basis of the grabbing action, so that the feeding process is changed from passive response to active adaptation, the linkage problem caused by positioning errors is eliminated, and the robustness of the detection system is enhanced.

In the present invention, preferably, a template matching step is provided between the image obtaining step and the detecting step, and includes invoking a standard template image corresponding to a current injection molding model, where a standard structural feature is preset in the standard template image, the standard structural feature corresponds to a preset detection type, extracting structural features from the collected image information, and performing similarity comparison with the standard structural feature, where a preset detection type corresponding to a standard structural feature with the highest similarity is used as the detection type of the current image information. According to the scheme, the template matching step is introduced, so that dynamic identification and adaptation of the detection type are realized, and the core problem of scheme mismatch in the detection of the multi-type injection molding is effectively solved. The method comprises the steps of calling a standard template image corresponding to the current injection molding model, ensuring that a reference standard is strictly consistent with an actual product model, avoiding deviation of a universal template in complex structure detection, providing accurate reference basis for subsequent comparison, presetting standard structural features in the standard template image, predefining geometric forms and visual attributes of key areas, enabling a system to distinguish different structure types such as slots and buckles, laying a structural foundation for selection of a detection strategy, enabling the standard structural features to correspond to preset detection types, directly associating the features with a specific detection scheme, ensuring accurate execution of the detection process aiming at the structural features, extracting the structural features from collected image information, capturing real visual performance of the actual injection molding under variables such as illumination and angles, providing reliable data input for similarity comparison, comparing the actual injection molding features with the standard structural features, effectively inhibiting misjudgment caused by background interference or slight deformation through quantifying the matching degree of the actual features and the standard structural features, enhancing the type recognition robustness, enabling the preset detection type corresponding to the structural features with the highest similarity to serve as the current detection type of the detection strategy, and achieving automatic adaptation of the automatic detection strategy, namely, achieving automatic adaptation and automatic detection of the automatic detection strategy.

In the present invention, preferably, a position correction step is provided between the template matching step and the detection step, including

Acquiring the actual position of the structural feature of the injection molding part in the image information, and acquiring the position of the standard structural feature in the standard template image and defining the position as an expected position;

Comparing the actual position with the expected position to obtain offset information;

Carrying out translation or rotation compensation on the detection area according to the offset information to realign the detection area with the actual position of the injection molding piece;

if the position correction fails, triggering a re-shooting process.

And in the position correction step, the accuracy of template matching is ensured by dynamically compensating the position deviation, so that the reliability of defect detection is improved. The method comprises the steps of obtaining the actual position of the structural feature of the injection molding part in image information, solving the problem of position uncertainty caused by part placement errors or mechanical arm positioning fluctuation, providing a real data basis for subsequent deviation analysis, obtaining the position of the standard structural feature in a standard template image and defining the position as an expected position, establishing a reference standard under an ideal state, enabling position comparison to have a clear basis, comparing the actual position with the expected position to obtain offset information, accurately identifying the details of position deviation through parameters such as quantitative translation amount, rotation angle and the like, providing key input for compensation operation, carrying out translation or rotation compensation on a detection area according to the offset information, enabling the detection area to be realigned with the actual position of the injection molding part, ensuring that the structural feature can be strictly corresponding when the template is matched, avoiding feature dislocation caused by position deviation, and remarkably improving accuracy of similarity comparison, triggering a re-shooting flow if the position correction fails, timely interrupting the error flow and re-acquiring the image, preventing the detection result based on the serious deviation from influencing subsequent judgment, and enhancing the fault tolerance and stability of a system. In the whole, the step effectively aims at unavoidable position fluctuation in the production environment through a closed-loop correction mechanism, and the robustness of the detection process is ensured.

The invention has the beneficial effects that:

According to the method, synchronous double-sided image acquisition is realized through the cooperative operation of the double mechanical arms, and the traditional overturning operation is avoided, so that the problems of efficiency, precision and reliability in double-sided detection are systematically solved. The method comprises the steps of feeding and conveying parts, namely, a feeding and conveying step, a detecting step, a discharging and conveying step, a detecting step and a discharging and conveying step, wherein the feeding and conveying step is used for stably conveying the parts to a detecting station through a feeding and conveying device and a clamping jaw, ensuring that the parts are accurate in position before entering a detecting flow, reducing positioning deviation caused by unstable conveying, laying a foundation for subsequent image acquisition, synchronously acquiring image information of preset points according to a preset sequence in the image acquisition step by an upper mechanical arm and a lower mechanical arm, realizing simultaneous detection of the upper surface and the lower surface, eliminating the problems of efficiency loss, secondary damage risk and dislocation of a detecting area caused by repeated conveying and turning, judging in real time by adopting a corresponding detecting scheme according to the detecting type of the detection points in the image information, rapidly outputting qualified or unqualified signals, ensuring that the defect identifying process is efficient and accurate, reducing the influence of image transmission delay on the beat of a production line, integrating all detecting results through an upper computer, and utilizing an equipment conveying belt and a discharging clamping jaw to automatically split qualified product and unqualified product, realizing instant response and efficient processing of the detecting results, and classification errors caused by manual intervention or complex mechanisms. The characteristics are tightly matched, the feeding conveying provides stable input to ensure the precision of a detection starting point, the image acquisition is synchronously completed to finish double-sided data acquisition, the interruption of a flow is avoided, the real-time analysis of the detection steps ensures rapid decision, the discharging conveying rapidly carries out classification to form a closed loop, the continuous detection flow without overturning is jointly constructed, and the overall efficiency and reliability of double-sided detection are obviously improved.

Drawings

FIG. 1 is a schematic diagram of the overall structure of a detection device in the present invention;

FIG. 2 is a schematic diagram showing the internal structure of the detecting device according to the present invention;

FIG. 3 is a schematic view of the structure of a detection station in the detection device of the present invention;

FIG. 4 is a simplified flow chart of the detection method of the present invention;

FIG. 5 is a schematic overall flow chart of the detection method of the present invention;

FIG. 6 is a flow chart of a slot inspection strategy in the inspection method of the present invention;

FIG. 7 is a schematic flow chart of a detecting strategy of a buckle spring in the detecting method of the present invention;

FIG. 8 is a schematic flow chart of a beam detection strategy in the detection method of the present invention;

FIG. 9 is a schematic diagram of the structure of a DHT framework of the present invention;

Figure 10 is a schematic structural diagram of a CMA framework in accordance with the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It will be understood that when an element is referred to as being "fixed to" another element, it can be directly on the other element or intervening elements may also be present. When a component is considered to be "connected" to another component, it can be directly connected to the other component or intervening components may also be present. When an element is referred to as being "disposed on" another element, it can be directly on the other element or intervening elements may also be present. The terms "vertical," "horizontal," "left," "right," and the like are used herein for illustrative purposes only.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

The present embodiment provides a defect detecting apparatus, please refer to fig. 1-3. In the drawing, the application is provided with a feeding clamping jaw and a discharging clamping jaw, a detection chamber is arranged between the feeding clamping jaw and the discharging clamping jaw, an upper mechanical arm and a lower mechanical arm are arranged in the detection chamber, wherein the upper mechanical arm and the lower mechanical arm are positioned at two sides of an injection molding piece to be detected, and an industrial camera and a stroboscope are respectively carried on the upper mechanical arm and the lower mechanical arm.

Referring to fig. 4 and 5, the injection molding defect detection method based on the cooperative operation of the dual mechanical arms provided by the invention comprises an S1 preparation step, an S2 feeding and conveying step, an S3 image acquisition step, an S4 template matching step, an S5 position correction step, an S6 detection step and an S7 blanking and conveying step, wherein the steps are closely connected to form a complete detection flow:

Before detection starts, firstly, a preparation step S1 is executed, namely, the model of the injection part to be detected is obtained, the detection equipment automatically adjusts the distance between the conveying belts of the equipment, the stop position and the preset image acquisition point position according to the model, and meanwhile, the corresponding detection parameters are adapted, so that the rapid switching detection of the injection parts of different models can be realized without complex hardware replacement or parameter reconstruction.

After the preparation step is finished, the S2 feeding and conveying step is carried out, namely the feeding and conveying device continuously conveys the injection part to be detected, when a sensor on the device detects that the injection part reaches a preset feeding station, the sensor sends a trigger signal, a feeding clamping claw responds to the signal and grabs the injection part to be detected, the injection part is stably placed on an equipment conveying belt, and the equipment conveying belt accurately conveys the injection part to a detection station, so that the initial position stability of subsequent detection is ensured.

After the detection station is reached, an S3 image acquisition step is executed, wherein the step is completed by the upper mechanical arm and the lower mechanical arm in a cooperative and synchronous way, the upper mechanical arm and the lower mechanical arm are pre-stored with motion tracks corresponding to the type of the injection molding to be detected, the upper mechanical arm and the lower mechanical arm are respectively moved to the optimal shooting position according to the preset tracks, the gesture is adjusted in real time through a position feedback signal, the shooting gesture is ensured to be kept stable with the distance of the surface of the injection molding, further image information of the preset point positions is respectively acquired, the synchronous acquisition of the images of the upper surface and the lower surface of the injection molding is realized, the complex process of the turnover re-detection after the traditional single-sided detection is avoided, and the detection efficiency is greatly improved.

In order to ensure accurate adaptation of detection types, an S4 template matching step is set after the image acquisition step, wherein standard template images corresponding to the current injection molding model are firstly called, standard structural features reflecting different detection types are preset in the standard template images, the geometric forms and visual attributes of different structures such as clamping grooves, through holes, buckles, elastic sheets and cross beams are defined by the features, the system extracts actual structural features from acquired image information, compares the actual structural features with the standard structural features in the standard template images in similarity, and finally the detection type corresponding to the standard structural feature with the highest similarity is determined as the detection type of the current image information, so that automatic matching of detection strategies is realized, and the pertinence of detection is improved.

The method comprises the steps of setting S5 a position correction step after a template matching step by considering position deviation possibly caused by injection molding placement errors or mechanical arm positioning fluctuation, firstly acquiring the actual position of structural features of the injection molding in acquired image information, simultaneously extracting the position of standard structural features in a standard template image and defining the position as an expected position, comparing the actual position with the expected position, calculating to obtain offset information such as translation amount, rotation angle and the like, carrying out translation or rotation compensation on a preset detection area according to the offset information to enable the detection area to be aligned with the actual structural position of the injection molding again accurately, and automatically triggering a re-shooting flow by a system to ensure that subsequent detection is carried out based on the accurately aligned image information if the deviation is found out to exceed a compensation range in the position correction process to cause correction failure.

After the position correction is completed, the system enters a core S6 detection step, and a corresponding detection scheme is called according to the detection type determined by template matching, wherein the detection scheme comprises a slot detection strategy aiming at a clamping slot or through hole area, a clamping buckle elastic sheet detection strategy aiming at a clamping buckle or elastic sheet structure and a beam detection strategy aiming at a beam area, and the problem of insufficient adaptability of a general scheme to complex structure defect identification is solved by customizing different detection strategies for different structural features.

Referring to fig. 6, the slot detection strategy is specifically implemented by precisely positioning the center position and boundary size of a card slot or a through hole in image information according to a preset structural template, further defining a corresponding slot detection area, ensuring that the detection area is strictly limited in a target structural range, avoiding background interference, dividing the slot detection area into a plurality of layers according to a direction perpendicular to an opening, simultaneously further subdividing each layer into a plurality of local areas, attaching the depth characteristics of holes in a dividing manner, precisely capturing the brightness variation along the depth direction of the hole, realizing defect positioning of microscopic layers, calculating the average brightness of each local area in each layer, taking the average brightness as the layer brightness value of the layer, inhibiting the influence of random noise and tiny reflection points through local average processing, comparing the layer brightness value of each layer with a preset normal brightness range, defining a layer exceeding the normal brightness range as a brightness abnormal layer, rapidly focusing a potential problem area, carrying out reflection interference identification on the brightness abnormal layer, carrying out brightness compensation treatment on the identified reflection area, and then carrying out brightness re-judging on the bottom layer caused by reflection of materials, and eliminating the situation that the abnormal brightness is not completely through the card slot or the boundary is completely through when the defect is completely judged to exist.

Referring to fig. 7, the specific implementation process of the buckle shrapnel detection strategy includes the steps of firstly, reading a buckle or shrapnel standard template corresponding to the current injection molding model, determining a buckle shrapnel detection area in image information according to the standard template, establishing accurate reference, dividing the buckle shrapnel detection area into a plurality of parts such as a head part, a connecting root part and a tail end according to structural characteristics of the buckle shrapnel detection area, respectively comparing the templates of the parts to obtain similarity between the parts and the standard template, executing edge extraction processing on the parts with the similarity lower than a preset threshold, further dividing the parts into a plurality of local areas, carrying out depth analysis on focusing problem areas, avoiding redundant calculation, calculating edge strength and edge trend consistency of the local areas, judging edge reliability through double indexes, effectively filtering interference such as reflection and noise, accurately calculating the actual edge position of the area and the offset of the corresponding edge position in the standard template when the edge reliability of the local area is insufficient, and judging that the buckle or shrapnel has fracture or deformation defects if the offset exceeds the preset allowable range.

Referring to fig. 8, the specific implementation process of the beam detection strategy includes determining the overall position of the beam in image information through boundary extraction and continuity analysis, laying a foundation for subsequent detection, dividing the beam into a plurality of continuous detection sections along the length direction of the beam, setting a certain range of preset overlapping amount between adjacent detection sections, eliminating defect omission risks of boundaries between sections, performing multi-level bright-dark separation processing on the inside of each detection section, dynamically adjusting a separation threshold according to local brightness gradient, adapting to scenes of light reflection unevenness and illumination change, accurately extracting a foreground region reflecting a beam structure, analyzing the area size and aspect ratio of the foreground region, forming quantized detection features, wherein the area size is used for identifying the material missing scale, the aspect ratio is used for capturing geometric distortion, sequentially executing multistage verification processes of foreground region quantity comparison, area ratio judgment and adjacent position foreground region continuity inspection, firstly removing random noise through quantity comparison, then confirming defect significance through area ratio judgment, and finally verifying spatial relevance through continuity inspection, and judging that if a plurality of continuous detection sections appear along the length direction of the beam, determining that a broken or material is missing exists.

After the detection step is finished, the system outputs a corresponding detection result, namely if the detection result of a certain detection point is unqualified, the detection image and the point information of the detection point are stored in a defect database, an unqualified signal is output to the upper computer, and if the detection result is qualified, only a qualified signal is output to the upper computer.

And S7, carrying out blanking conveying step, namely conveying the injection molded part subjected to detection to a blanking station by using a conveying belt of equipment, collecting detection results of all detection points by using an upper computer to carry out final quality judgment, judging that the injection molded part is a defective product if any detection point outputs a defective signal, enabling the injection molded part to directly flow out along with the conveying belt of the equipment, judging that the injection molded part is a defective product if all detection points output a defective signal, grabbing the injection molded part by using a blanking clamping claw, and placing the injection molded part in a defective product area to complete the whole detection flow.

Referring to fig. 9 and 10, in this embodiment, the injection molding to be detected is a DHT model and a CMA model of an automobile front end frame, and the detection device includes a feeding conveying device, a device conveying belt, an upper mechanical arm, a lower mechanical arm, an industrial camera, a sensor, a feeding clamping jaw, an upper computer and a defect database, wherein industrial-level high-resolution cameras are mounted at the tail ends of the upper mechanical arm and the lower mechanical arm, and the sensor adopts an opposite-injection photoelectric sensor for detecting an in-place state of the injection molding.

Before detection starts, the preparation steps are executed, wherein an operator inputs the model of an injection molding to be detected, such as a DHT model, through an upper computer, after equipment receives model information, the distance between equipment conveying belts is automatically adjusted to adapt to the width of the DHT model injection molding, meanwhile, the stop position of the equipment conveying belts is adjusted, the injection molding is ensured to be positioned at a preset central position when conveyed to a detection station, the preset movement track and the image acquisition point position of an upper mechanical arm and a lower mechanical arm are synchronously switched to parameters corresponding to the DHT model, the grabbing sizes of a feeding clamping claw and a discharging clamping claw are automatically adapted to the structural size of the DHT model injection molding, and the whole preparation process is free from manually adjusting hardware or reconstruction parameters, so that model switching is rapidly completed.

After the preparation step is finished, starting a feeding conveying device, moving the injection molding piece of the DHT model to be detected along with the feeding conveying device, sending a trigger signal to a control system by a sensor when an opposite-irradiation photoelectric sensor on the device detects that the injection molding piece reaches a feeding station, driving a feeding clamping claw to act by the control system, accurately grabbing the injection molding piece and stably placing the injection molding piece on an equipment conveying belt, conveying the injection molding piece to a detection station by the equipment conveying belt at a preset speed, stopping running by the equipment conveying belt after the injection molding piece reaches the detection station, and waiting for image acquisition.

The image acquisition step is started, the upper mechanical arm and the lower mechanical arm synchronously move to the optimal shooting positions of the upper surface and the lower surface of the injection molding piece according to the preset motion trail corresponding to the DHT model, a position feedback module on the mechanical arm acquires gesture and distance information in real time, the distance between a camera and the surface of the injection molding piece is ensured to be kept constant, the defect of insufficient image definition caused by distance fluctuation is avoided, then, the industrial cameras on the upper mechanical arm and the lower mechanical arm synchronously shoot, the image information of a preset point position is acquired, and the image information is transmitted to a system processing unit in real time.

And the system extracts the actual structural features of each structure from the acquired image information through an image feature extraction algorithm, compares the actual structural features with the standard structural features by adopting a normalized cross-correlation matching algorithm, and finally takes a detection type corresponding to the standard structural feature with highest similarity as a clamping groove structure, and distributes a corresponding slot detection strategy to a corresponding region in the current image information.

After template matching is completed, a position correction step is executed, wherein a system extracts pixel coordinates of an actual structural feature in an acquired image as an actual position, and simultaneously extracts pixel coordinates of a standard structural feature in a standard template image as an expected position, a translation amount and a rotation angle of the actual position relative to the expected position are calculated through coordinate comparison, translation and rotation compensation are carried out on a preset detection area according to the translation amount and the rotation angle, so that the detection area is accurately aligned with the actual structural position of an injection molding piece, in the embodiment, if the calculated translation amount exceeds 5 pixels or the rotation angle exceeds 3 degrees, correction failure is judged, the system triggers a re-shooting instruction, and an upper mechanical arm and a lower mechanical arm execute an image acquisition flow again.

After the position correction is successful, the detection step is entered, the system calls a slot detection strategy, firstly, the slot detection area is defined in an image and divided into 3 layers according to the direction perpendicular to an opening, each layer is divided into 8 local areas, after the average brightness of each local area of each layer is calculated, the average brightness of the bottom layer is compared with a normal brightness range, the average brightness of the bottom layer is found to exceed the normal range, the reflective area of the bottom layer is further identified and brightness compensation is carried out, after compensation, the brightness of the bottom layer is recalculated and still is abnormal, and a fracture exists at the boundary of the bottom layer, therefore, the card slot is judged to have a non-through defect, the system stores the detection image and the corresponding point number into a defect database, and outputs a disqualification signal to an upper computer, the buckle structure area is called, the buckle elastic sheet detection strategy is divided into 3 parts according to the structure characteristics, the template is compared, the similarity of the connection root is found to be low, the part is extracted with the edge and divided into 5 local areas, the edge strength is obtained by calculation, the deviation quantity of the edge of the template is further calculated, the edge is larger than the preset range, the fracture quantity of the bottom layer is still is calculated, the fracture quantity of the bottom layer is still is abnormal, and the fracture quantity of the bottom layer is observed, and the fracture quantity is judged, the defect exists at the position of the bottom layer is detected, and the defect is continuously, and the defect is detected by the same, and the position of the adjacent segment is detected by the adjacent segment with the cross beam with the position detection section, and the defect is detected by the position detection section with the position detection section and has a bright-level and the same, and the position detection section and has a position-level and the position detection stage.

After all the detection points are detected, the equipment conveyer belt is started to convey the injection molding part to the blanking station, the upper computer gathers the results of all the detection points, and the injection molding part is judged to be an unqualified product due to a plurality of unqualified signals, and directly flows out along with the equipment conveyer belt, if the detection is the CMA type injection molding part, the detection flow is consistent with the DHT type, the CMA type is only required to be input in a preparation step, and the detection can be completed by automatically adjusting related parameters by the equipment without additional hardware adjustment or algorithm reconstruction.

Based on the injection molding detection method provided by the invention, industrial application verification for three months is carried out on the production line of a certain injection molding manufacturing enterprise. The test environment completely simulates actual production conditions, and the detection results are statistically analyzed by carrying out full-flow tracking detection on continuously produced injection molding pieces of the front end frame of the automobile. Experimental data show that the system has excellent detection performance, namely that under a standard detection mode, the average time for finishing all image acquisition and detection analysis of 39 key detection points of a single injection molding part is only within 2 seconds, and the beat requirement of a production line is completely met. More significantly, the system shows high stability in the continuous operation process, and the false detection rate is always kept at a low level through statistics, so that the standard requirement of enterprises is met. The quantitative indexes fully verify the superior performance of the invention in terms of detection efficiency, accuracy and stability, and provide a reliable technical solution for online quality monitoring in an intelligent manufacturing environment.

The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the present invention may occur to one skilled in the art without departing from the principles of the present invention and are intended to be within the scope of the present invention.

Claims

1. The injection molding defect detection method based on the cooperative operation of the double mechanical arms is characterized by comprising the following steps of

Feeding and conveying, namely conveying the injection molded parts to a detection station by using an equipment conveying belt;

Template matching, namely acquiring a detection type based on image information;

A position correction step, namely acquiring the actual position of the structural feature of the injection molding in the image information, acquiring the position of the standard structural feature in the standard template image, defining the position as an expected position, comparing the actual position with the expected position to acquire offset information, carrying out translation or rotation compensation on a detection area according to the offset information, realigning the detection area with the actual position of the injection molding, and triggering a re-shooting process if the position correction fails;

A detection step, namely acquiring a detection result by adopting a corresponding detection scheme according to the detection type of the detection point in the image information;

And in the blanking conveying step, the injection molding part is conveyed to a blanking station by the equipment conveying belt, and the upper computer carries out quality judgment on the injection molding part according to the detection results of all the detection points.

2. The method for detecting defects of injection molding based on cooperative operation of two mechanical arms according to claim 1, wherein the detection scheme is used for identifying breakage, shortage, non-penetration, deformation and other appearance defects, and comprises a slot detection strategy, a buckle shrapnel detection strategy and a beam detection strategy, wherein the slot detection strategy executes a brightness detection strategy aiming at a clamping slot or a through hole area, the buckle shrapnel detection strategy executes a template matching and edge analysis strategy aiming at a buckle or shrapnel structure, and the beam detection strategy executes a connectivity analysis strategy aiming at a beam area.

3. The method for detecting defects of injection molding based on cooperative operation of two mechanical arms according to claim 2, wherein the slot detection strategy comprises:

4. The method for detecting defects of injection molding based on cooperative operation of two mechanical arms according to claim 2, wherein the snap spring detection strategy comprises:

5. The method for detecting defects of injection molding based on cooperative operation of double mechanical arms according to claim 2, wherein the beam detection strategy comprises:

Dividing a plurality of continuous detection sections along the length direction of the cross beam, and arranging a preset overlap amount between adjacent detection sections, wherein the preset overlap represents the overlap length between the adjacent detection sections;

6. The method for detecting defects of injection molding based on collaborative operation of two mechanical arms according to claim 1, wherein the image acquisition step is completed by an upper mechanical arm and a lower mechanical arm synchronously, wherein the upper mechanical arm and the lower mechanical arm respectively move to shooting positions according to preset motion tracks, and shooting postures and distances are ensured to be stable through position feedback signals.

7. The method for detecting the defects of the injection molding based on the cooperative operation of the double mechanical arms according to claim 1, wherein a preparation step is further arranged before the feeding and conveying step, the preparation step comprises the steps of obtaining the model of the injection molding part to be detected, and the detection equipment adjusts the distance between the equipment conveying belts, the stop position and the preset point position according to the model.

8. The method for detecting defects of injection molding based on cooperative operation of double mechanical arms according to claim 1, wherein the feeding and conveying step comprises the step of grabbing the injection molding to be detected by the feeding clamping claw and arranging the injection molding to be detected on the equipment conveying belt when the sensor on the feeding and conveying device detects that the injection molding arrives.

9. The method for detecting the defects of the injection molding based on the collaborative operation of the double mechanical arms according to claim 1, wherein the template matching step comprises the steps of calling a standard template image corresponding to the current injection molding model, wherein standard structural features are preset in the standard template image, the standard structural features correspond to preset detection types, the structural features are extracted from the acquired image information and are compared with the standard structural features in similarity, and the preset detection type corresponding to the standard structural feature with the highest similarity is used as the detection type of the current image information.