CN109900711A - Workpiece, defect detection method based on machine vision - Google Patents

Workpiece, defect detection method based on machine vision Download PDF

Info

Publication number
CN109900711A
CN109900711A CN201910264717.XA CN201910264717A CN109900711A CN 109900711 A CN109900711 A CN 109900711A CN 201910264717 A CN201910264717 A CN 201910264717A CN 109900711 A CN109900711 A CN 109900711A
Authority
CN
China
Prior art keywords
workpiece
pixel
sub
segmentation
defect
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201910264717.XA
Other languages
Chinese (zh)
Inventor
耿磊
魏全生
肖志涛
吴骏
张芳
李文科
刘彦北
王雯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tiangong University
Original Assignee
Tianjin Polytechnic University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin Polytechnic University filed Critical Tianjin Polytechnic University
Priority to CN201910264717.XA priority Critical patent/CN109900711A/en
Publication of CN109900711A publication Critical patent/CN109900711A/en
Pending legal-status Critical Current

Links

Landscapes

  • Image Processing (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

本发明提供了一种基于机器视觉的工件缺陷检测方法。该方法首先采集法兰盘式工件的图像，对相机进行标定，获取标定误差，然后对工件轮廓进行亚像素边缘信息的提取，计算出拟合边缘到工件轮廓的距离，并通过比较该距离是否大于所给阈值判别工件外轮廓的破损情况，最后针对工件表面纹理复杂影响工件表面划痕与锈蚀分割的问题，采用基于像素分层采样的PixelNet卷积神经网络对表面缺陷进行分割。结果表明，本发明可以准确地检测出工件的外形缺陷和表面缺陷，并提高了算法鲁棒性。

The invention provides a workpiece defect detection method based on machine vision. The method first collects the image of the flanged workpiece, calibrates the camera, obtains the calibration error, then extracts the sub-pixel edge information of the workpiece contour, calculates the distance from the fitted edge to the workpiece contour, and compares whether the distance is The damage of the outer contour of the workpiece is judged when the threshold is greater than the given threshold. Finally, the PixelNet convolutional neural network based on pixel layered sampling is used to segment the surface defects for the problem that the complex surface texture of the workpiece affects the segmentation of the workpiece surface scratches and rust. The results show that the invention can accurately detect the shape defects and surface defects of the workpiece, and improve the robustness of the algorithm.

Description

Workpiece, defect detection method based on machine vision
Technical field
The workpiece, defect detection method based on machine vision that the present invention relates to a kind of is using image to flange plate type workpiece When processing technique is detected, detection accuracy and detection efficiency are improved.
Background technique
The computer speed of service is significantly improved from the 1960s, and the appearance of simultaneous CCD technology is based on The defect detecting technique of machine vision starts to be used widely in industrial production line, such as machinery, electronics, printing, weaving row Industry improves product quality and production efficiency by advanced detection technique.A kind of side of the Machine Vision Detection as non-destructive testing Method, it obtains clearly object under test image by linear array or area array cameras, carries out image procossing completion target by computer and lacks Sunken real-time detection.
With the fast development of China's manufacturing industry, the requirement for product quality is also higher and higher, is based on machine in recent years The defects detection theoretical research of vision is constantly mature.Zhou Shan Min et al. obtains metal surface by the way of different angle illumination Multiplex images, defect characteristic has significant difference under different light angles, and the variation between multiple image provides more lack Characteristic information is fallen into, the detection to cracks of metal surface is realized by the related information excavated between multiplex images.This method needs not The disconnected light angle that changes shoots multiple image, but workpiece is in mobile state in the present invention, carries out multi-angle illumination and not only grasps Make cumbersome and can not accomplish real-time detection and sorting to workpiece.Huang Liuqian et al. is asked for the detection of stamping parts appearance defect Topic carries out closed operation to realize the filling of shape damage location, then by the background after closed operation to the background area of extraction Region makes the difference with original image background area, and then the breakage of stamping parts is judged according to size of the difference.Due to workpiece breakage size Difference, therefore the selection of structural elements can not unify when carrying out closed operation to workpiece, to influence the precision of appearance defect detection. Li Yongjing et al., which is proposed, determines workpiece for measurement position based on the matched method of shape template, passes through shape extraction and dynamic threshold Method workpiece configurations defect is split.Since it is using obtaining workpiece image by the way of illuminating, the method is not Detection suitable for the present invention to workpiece configurations defect.Huang Jingwei et al. proposes the image alignment algorithm based on contour feature, To realize the quick alignment of hardware workpiece and template image, and by judging that workpiece for measurement image and standard workpiece image grayscale are poor The size detection Surface Flaw of value.This method is poor to fine scratches and lesser defect Segmentation effect, is unable to satisfy this Real-time detection of the invention to Surface Flaw.
In conclusion there is an urgent need to propose a kind of accuracy height and the stronger workpiece configurations defect of practicability and surface at present The detection method of defect.
Summary of the invention
The purpose of the present invention is overcoming the above-mentioned deficiency of existing method, a kind of workpiece, defect based on machine vision is proposed Detection method, this method may be implemented to greatly improve workpiece configurations defect and surface defect measurement accuracy.For this purpose, of the invention It adopts the following technical scheme that:
Step 1: the image of flange plate type workpiece in acquisition reality；
Step 2: camera being demarcated using Zhang Shi standardization, the parameter and pose of camera are obtained, according to calibration result Obtain the calibrated error of measuring system；
Step 3: image filtering being carried out to workpiece image, extracts area-of-interest；
Step 4: the detection of workpiece pixel edge is completed using Canny operator first, then using based on Gray Moment The sub-pixel edge of method extraction workpiece image；
Step 5: the mass center of concentric holes and via hole on workpiece being obtained using the method for circle fitting, according to the think of of least square method Think, the edge for approaching workpiece outer profile and each via hole is gone to circle；
Step 6: using the defect inspection method based on edge pixel distance, the sub-pix calculated in workpiece actual profile is sat Distance of the punctuate to fitting circle radial direction；
Step 7: surface defect being split using the PixelNet convolutional neural networks based on pixel stratified sampling, is obtained To scratch and corrosion region.
Compared with prior art, the beneficial effects of the present invention are:
For the low problem of workpiece configurations defect and surface defects detection difficulty and detection efficiency, the invention proposes be based on The defect inspection method of machine vision detects the appearance defect of workpiece by Edge Distance method, in conjunction with pixel stratified sampling PixelNet convolutional neural networks are split detection to Surface Flaw.
Compared to traditional detection method, the present invention has many advantages, such as that non-contact, precision is high, adaptability is good, by worker from numerous It is freed in superfluous detection work, greatly improves detection efficiency.The present invention is directed to workpiece outer profile damaged area shape The problem of its segmentation is with identification is influenced with size, the appearance defect detection method based on edge of work distance is proposed, to difference The workpiece of damaged degree has carried out test experience, and the recognition accuracy of outer profile breakage workpiece demonstrates the algorithm up to 100% Validity.In addition, aiming at the problem that workpiece surface texture complex effects workpiece surface scratch and corrosion are divided, using based on pixel The PixelNet convolutional neural networks of stratified sampling are split surface defect, the experimental results showed that the method for the present invention can be right Surface Flaw is effectively divided, the average friendship of segmentation result and ratio is up to 92.3%, therefore, method proposed by the invention It may be implemented to workpiece scratch, corrosion, effective detection of outer profile breakage workpiece, and detection accuracy is high.
Detailed description of the invention
Fig. 1 overall framework schematic diagram；
Fig. 2 flange plate type workpiece image；
Fig. 3 pixel edge extracts result；
Fig. 4 extracts result based on the sub-pixel edge of Gray Moment；
Fig. 5 workpiece breakage position view；
Fig. 6 workpiece breakage position enlarged drawing；
Fig. 7 (a) is accuracy rate change curve, (b) is Loss value change curve；
Fig. 8 corrodes segmentation result；
Fig. 9 scratch segmentation result.
Specific embodiment
The embodiment of the present invention is described in further detail with reference to the accompanying drawing.
Overall framework flow diagram of the invention is as shown in Figure 1.Firstly, acquisition flange disk workpiece image, using Family name's standardization demarcates camera, and then carries out distortion correction to workpiece image；Then it uses using Gaussian filter to figure As being smoothed, convolution operation is carried out using 5 × 5 Gaussian kernels that standard deviation is 1, extracts area-of-interest；It is respectively adopted Canny algorithm, Sobel algorithm, Roberts algorithm and Prewitt algorithm carry out pixel edge detection to image, more various Algorithm detection as a result, the best Canny algorithm of selective extraction effect；Using the Sub-pixel Edge Detection based on Gray Moment Extract the sub-pixel edge of workpiece image；Finally, the mass center of concentric holes and via hole above workpiece is obtained using the method for circle fitting, According to the thought of least square method, the edge for approaching workpiece outer profile and each hole is gone to circle；Workpiece is calculated by Edge Distance Appearance defect；Surface defect is split using the PixelNet convolutional neural networks based on pixel stratified sampling, is obtained The corrosion of workpiece surface and scratch.
1. experimental subjects
Visual field size of the invention is 120mm × 90mm, and selected industrial camera resolution ratio is 1280 × 960, can be calculated The corresponding actual physics distance of each pixel is about 0.09375mm.It is tested using flange plate type workpiece, workpiece configurations are damaged Degree arc length size μ as corresponding to damage location is indicated, tests μ≤1mm respectively, tri- kinds of 2mm and μ >=2mm of μ < of 1mm < The workpiece of different size, Surface Flaw include corrosion and scratch.
2. pixel edge detects
Canny algorithm, Sobel algorithm, Roberts algorithm, Prewitt algorithm is respectively adopted, Pixel-level side is carried out to image Edge detection.Though by Sobel and Roberts edge detection algorithm it can be seen from the extraction result of four kinds of edge detection operators in Fig. 3 The edge of workpiece single pixel can be so obtained, but sweep is poor, there are jagged edges, are unfavorable for the accurate of sub-pixel edge It extracts.For Prewitt operator edge extracting result there are false edge, detection effect is poor.Canny edge detection algorithm is to workpiece Edge extracting is more accurate, while edge-smoothing degree is higher, and the present invention need to measure workpiece size, to edge positioning Required precision is higher, therefore the present invention selects Canny algorithm to the edge extracting of workpiece progress Pixel-level.The visual field size of system For 120mm × 90mm, selected industrial camera resolution ratio is 1280 × 960, can calculate the corresponding actual physics of each pixel away from From about 0.09375mm.There may be the error of 1 pixel between two o'clock in actual measurement, since target of the present invention is examined It surveys precision and is less than 0.1mm, pixel edge is affected to detection accuracy, is unable to satisfy high-precision measurement request.
3. sub-pixel edge extracts
The pixel edge known to 2 is affected to workpiece sensing precision, is unable to satisfy the measurement request of this system.Guarantee In the case that visual field is constant, improving measurement accuracy most straightforward approach can be used the industrial camera of higher resolution to reduce pixel Equivalent, but data volume can be made to increase simultaneously, it is unfavorable for real-time detection.And sub-pixel edge detection can reach 1/n pixel, be equivalent to Measurement accuracy is improved n times.The measurement demand of higher precision can be completed in the case where guaranteeing camera and the constant visual field.Therefore The present invention uses Canny algorithm to complete the detection of workpiece pixel edge first, is then examined again using the edge based on Gray Moment Survey method.Workpiece is detected based on the sub-pixel edge of Gray Moment, sub-pixel edge extracts result as shown in figure 4, using ash When spending moments method extraction workpiece subpixel coordinates, arithmetic accuracy is higher, while sub-pixel edge is smooth, is more suitable for workpiece Asia picture The positioning at plain edge.
4. the defect inspection method based on Edge Distance
The common appearance defect of workpiece is caused mainly due to the incompleteness of edge of work part, on the image show as workpiece Actual edge profile and fitting circle between there are the deviations of certain distance.As shown in figure 5, wherein red edge is the reality of workpiece Border profile, yellow edge are the workpiece profile that fitting obtains.If the collection of sub-pix point is combined into A in workpiece actual profilei(Xi, Yi), I ∈ (1,2,3 ... N), it is known that the center of circle of workpiece fitting circle is O (m, n), by center of circle O (m, n) and AiDetermining linear equation can table It is shown as:Abbreviation obtains: (Xi-m)y+(n-Yi)x-nXi+mYi=0, if the radius of workpiece fitting circle is R,Two coordinate points are calculated, wherein with point AiIt is apart from the smallest point Coordinate points in fitting circle, are denoted as Bi(Pi, Qi).Then sub-pix point A in workpiece actual edgeiAlong fitting circle radial direction to quasi- Close the distance d of profileiIt can pass throughIt finds out.Workpiece is at intact unbroken position, actual wheel Sub-pix point A on exterior featureiWith the radial direction distance d of fitting circleiMuch smaller than the distance of workpiece breakage position.Distance threshold τ is set, Work as diThe sub-pix point A in workpiece actual profile at this time is recorded when > τiThat is workpiece breakage position, thus according to threshold tau is met Coordinate points AiNumber { 0, n } can determine that whether current workpiece is damaged workpiece.
In practical application, it is contemplated that sub-pix point number more (N > 2000) influences the operation speed of algorithm on workpiece profile Degree, the present invention is to sub-pix point set AiIt is sampled in 360 ° relative to workpiece mass center, i.e., by Ai360 parts of equal parts are carried out, A coordinate points ξ is taken out at random in per sub-pix space of points ξ oncei,Constitute new work Part contour pixel point set A 'j, thenThen pass through the pixel collection A ' after samplingjInto Row diCalculating, compared to directly by former ensemble space AiIt is calculated, operation times are greatly decreased, while sub-pix point set A 'j Largely remain former set AiFeature, that is, workpiece profile information, and the differentiation of workpiece breakage situation not will cause It influences, Fig. 6 is the partial enlarged view of workpiece breakage position, A 'jFor the coordinate set after workpiece profile sub-pix point sampling, BiFor Coordinate set in fitting circle corresponding with damage location sub-pix point.Green line segment is then distance of the damage location to fitting circle di.For the robustness for verifying the algorithm, the present invention carries out test experience to the workpiece under different damaged degree, to 100 differences The workpiece of damaged degree is detected, and the recognition accuracy of algorithm is 100% under different damaged degree.
5. the PixelNet convolutional neural networks based on pixel stratified sampling are split surface defect
The present invention is acquired 3000 altogether by vision system and tested with scratch with the flange workpiece for corroding defect.It adopts Region of interesting extraction is carried out to workpiece with traditional image partition method, by workpiece and background separation.By PS software with hand The mode of work mark completes the mark to scratch on workpiece and corrosion.Meanwhile in order to which facilitating for supervised learning training will be on workpiece Scratch be indicated from corrosion with different characteristic values, so that mark figure is converted to index map, the wherein color mark of background Be denoted as (0,0,0), characteristic value 0, the color mark of scratch is (255,0,0), characteristic value 1, the color mark of corrosion be (0, 0,255), characteristic value 2.
Select 2400 pictures as training set from data set, 300 collect as verifying, and 300 are used as test set pair PixelNet network is trained and tests.The training of parted pattern of the present invention uses depth under 7 operating system of Windows The Matlab development interface of learning framework Caffe.In hardware environment, processor is Intel Xeon E5-2625, and video card is NVIDIA GTX1080Ti, video memory size are 11G.Training process is using 24 figures as an iteration step-length, wherein initial learning rate It is set as 0.01, as the increase of exercise wheel number finally decays to 0.0001, the number of iterations is 100,000 times.Standard in training process True rate variation is with penalty values variation as shown in Fig. 7 (a) and Fig. 7 (b).Segmentation result such as Fig. 8 and Fig. 9 institute to corrosion with scratch Show.By Fig. 7 (a) and Fig. 7 (b), the defect based on PixelNet convolutional neural networks that the present invention uses can be intuitively seen Dividing method can complete the accurate segmentation to Surface Flaw.It can accomplish have simultaneously for the different scratch of depth degree Effect is extracted, algorithm robustness with higher.Further to evaluate the method for the present invention to the segmentation performance of workpiece, defect, the present invention The standard evaluation index MIoU of selection semantic segmentation evaluates test result.The model generated using training is on test set It is tested, acquiring MIoU is 93.2%, accuracy rate with higher, can satisfy industry to the detection demand of workpiece, defect.
By judging that the Pixel Information of workpiece segmentation result can determine that workpieces are other, i.e. output result contains red pixel Can be identified as have the workpiece of scratch defects, containing blue pixel then for exist corrosion workpiece.Theoretically qualified workpiece Output result there was only black pixel value, but the case where will appear erroneous segmentation in practical cutting procedure, so that qualified work Part can also have less red or blue pixel value.Setting area threshold alpha is then determined as defect work when dividing defect and being greater than α Otherwise part is qualified workpiece, α value of the present invention is 10 pixels.According to actually detected requirement, the value of α can be finely tuned.
Workpiece, defect detection method proposed by the present invention based on machine vision, achievable shape different damaged degree and table The workpiece sensing of face existing defects, while differentiating that accuracy height, strong robustness can meet the needs of industrial detection.
The foregoing is merely of the invention preferably to apply example, is not intended to limit the scope of the present invention, it should be understood that this Invention is not limited to existing scheme as described herein, and the purpose of these implementations description is to help those skilled in the art The member practice present invention.Any those of skill in the art are easy to carry out without departing from the spirit and scope of the present invention Further improve and perfect, therefore the present invention is only limited by the content and range of the claims in the present invention, is intended to contain Covering all includes alternative and equivalent program in the spirit and scope of the invention being defined by the appended claims.

Claims (3)

1.一种基于机器视觉的的工件缺陷检测方法，包括下列步骤：1. A workpiece defect detection method based on machine vision, comprising the following steps: 步骤1：采集现实中法兰盘式工件的图像；Step 1: Collect the image of the flanged workpiece in reality; 步骤2：对相机进行标定，获取测量系统的标定误差；Step 2: Calibrate the camera to obtain the calibration error of the measurement system; 步骤3：对图像进行预处理提取出感兴趣区域；Step 3: Preprocess the image to extract the region of interest; 步骤4：采用亚像素边缘检测算法得到图像的亚像素边缘；Step 4: use the sub-pixel edge detection algorithm to obtain the sub-pixel edge of the image; 步骤5：通过拟合圆的方式得到法兰盘式工件的理论外边缘；Step 5: Obtain the theoretical outer edge of the flanged workpiece by fitting a circle; 步骤6：采用基于边缘距离的方法检测出工件的外形缺陷；Step 6: Use the edge distance-based method to detect the shape defects of the workpiece; 步骤7：采用基于像素分层采样的PixelNet卷积神经网络对表面缺陷进行分割。Step 7: Segment surface defects using PixelNet convolutional neural network based on pixel layered sampling. 2.根据权利要求1所述的基于机器视觉的工件缺陷检测方法，其特征在于，步骤6中，工件在完整无破损位置处，其实际轮廓上亚像素点Ai与拟合圆的半径方向距离di远小于工件破损位置的距离，设置距离阈值τ，当di＞τ时记录此时工件实际轮廓上的亚像素点Ai即工件破损位置，从而根据符合阈值τ的坐标点Ai的个数{0，n}即可判定当前工件是否为破损工件，实际应用中，考虑到工件轮廓上亚像素点个数较多(N＞2000)影响算法的运算速度，本发明对亚像素点集合Ai在相对于工件质心的360°内进行采样，即将Ai进行360份等分，在每一度的亚像素点空间ξ内随机取出一坐标点构成新的工件轮廓像素点集合A′j，则j∈(1，2，3…360)，然后通过采样后的像素点集合A′j进行di的计算，相比直接由原集合空间Ai进行计算，运算次数大幅减少，同时亚像素点集合A′j极大程度上保留了原集合Ai的特征即工件的轮廓信息，并对工件破损情况的判别不会造成影响。2. The workpiece defect detection method based on machine vision according to claim 1, is characterized in that, in step 6, the workpiece is at the complete and unbroken position, the sub-pixel point A i and the radial direction of the fitting circle on its actual outline The distance d i is much smaller than the distance of the workpiece damage position, and the distance threshold τ is set. When d i >τ, the sub-pixel point A i on the actual contour of the workpiece at this time is recorded, that is, the workpiece damage position, so that according to the coordinate point A i that meets the threshold τ The number of {0, n} can determine whether the current workpiece is a damaged workpiece. In practical applications, considering that the number of sub-pixel points on the workpiece contour is large (N>2000), which affects the operation speed of the algorithm, the present invention is suitable for sub-pixel points. The point set A i is sampled within 360° relative to the center of mass of the workpiece, that is, A i is divided into 360 equal parts, and a coordinate point is randomly selected in the sub-pixel point space ξ of each degree Constitute a new workpiece contour pixel set A' j , then j∈(1, 2, 3...360), and then calculate d i through the sampled pixel set A' j . Compared with the direct calculation from the original set space A i , the number of operations is greatly reduced, and the sub-pixel points The set A'j largely retains the feature of the original set A i , that is, the contour information of the workpiece, and will not affect the judgment of the damage of the workpiece. 3.根据权利要求1所述的基于机器视觉的工件缺陷检测方法，其特征在于，步骤7中，由于工件表面纹理较多，同时缺陷深浅程度不一且形状各异，传统分割算法无法对缺陷进行有效的分割，本发明首次采用PixelNet卷积神经网络完成对工件表面缺陷的分割，采用传统的图像分割方法对工件进行感兴趣区域提取，将工件与背景分离，借助PS软件以手工标注的方式完成对工件上划痕与锈蚀的标注，同时，为了监督学习训练的方便将工件上的划痕与锈蚀用不同的特征值进行表示，从而将标注图转换为索引图，其中背景的颜色标记为(0，0，0)，特征值为0，划痕的颜色标记为(255，0，0)，特征值为1，锈蚀的颜色标记为(0，0，255)，特征值为2，通过判断工件分割结果的像素信息即可确定工件类别，即输出结果含有红色像素的可确定为具有划痕缺陷的工件、含有蓝色像素的则为存在锈蚀的工件，理论上合格工件的输出结果只有黑色像素值，但在实际分割过程中会出现错误分割的情况，以至于合格工件也会存在较少红色或蓝色像素值，设置面积阈值α，当分割缺陷大于α时则判定为缺陷工件，否则为合格工件，本发明α取值为10个像素。3. The workpiece defect detection method based on machine vision according to claim 1, is characterized in that, in step 7, because the workpiece surface texture is many, and the defect depth is different and the shape is different simultaneously, the traditional segmentation algorithm cannot detect the defect. For effective segmentation, the present invention uses the PixelNet convolutional neural network for the first time to complete the segmentation of the surface defects of the workpiece, uses the traditional image segmentation method to extract the region of interest on the workpiece, separates the workpiece from the background, and uses PS software to manually mark the method. The marking of scratches and rust on the workpiece is completed. At the same time, for the convenience of supervised learning and training, the scratches and rust on the workpiece are represented by different eigenvalues, so as to convert the annotation map into an index map, in which the color of the background is marked as (0, 0, 0), the eigenvalue is 0, the scratch color is marked (255, 0, 0), the eigenvalue is 1, the rust color is (0, 0, 255), and the eigenvalue is 2, The workpiece category can be determined by judging the pixel information of the workpiece segmentation result, that is, the output result containing red pixels can be determined as a workpiece with scratch defects, and a workpiece containing blue pixels is a workpiece with rust. In theory, the output result of qualified workpieces There are only black pixel values, but in the actual segmentation process, there will be incorrect segmentation, so that there will be fewer red or blue pixel values in qualified workpieces. The area threshold α is set, and when the segmentation defect is greater than α, it is determined as a defective workpiece , otherwise it is a qualified workpiece, and the value of α in the present invention is 10 pixels.
CN201910264717.XA 2019-04-02 2019-04-02 Workpiece, defect detection method based on machine vision Pending CN109900711A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910264717.XA CN109900711A (en) 2019-04-02 2019-04-02 Workpiece, defect detection method based on machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910264717.XA CN109900711A (en) 2019-04-02 2019-04-02 Workpiece, defect detection method based on machine vision

Publications (1)

Publication Number Publication Date
CN109900711A true CN109900711A (en) 2019-06-18

Family

ID=66955230

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910264717.XA Pending CN109900711A (en) 2019-04-02 2019-04-02 Workpiece, defect detection method based on machine vision

Country Status (1)

Country Link
CN (1) CN109900711A (en)

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443791A (en) * 2019-08-02 2019-11-12 西安工程大学 A kind of workpiece inspection method and its detection device based on deep learning network
CN110441318A (en) * 2019-08-22 2019-11-12 郑州大学 A kind of chemical fibre spinneret hole defect inspection method based on machine vision
CN110579184A (en) * 2019-10-24 2019-12-17 深圳宇骏视觉智能科技有限公司 A product appearance online detection device and its use method
CN110602462A (en) * 2019-09-27 2019-12-20 南京工程学院 Industrial image detection device and method based on AI
CN110687120A (en) * 2019-09-18 2020-01-14 浙江工商大学 Flange appearance quality inspection system
CN111062915A (en) * 2019-12-03 2020-04-24 浙江工业大学 A real-time steel pipe defect detection method based on improved YOLOv3 model
CN111189854A (en) * 2020-04-13 2020-05-22 征图新视(江苏)科技股份有限公司 Defect layering detection method of automatic glass cover plate detection system
CN111784662A (en) * 2020-06-29 2020-10-16 深圳至峰精密制造有限公司 Workpiece identification method, device, computer equipment and storage medium
CN112508939A (en) * 2020-12-22 2021-03-16 郑州金惠计算机系统工程有限公司 Flange surface defect detection method, system and equipment
CN112700440A (en) * 2021-01-18 2021-04-23 上海闻泰信息技术有限公司 Object defect detection method and device, computer equipment and storage medium
CN113588681A (en) * 2021-07-30 2021-11-02 广东韶钢松山股份有限公司 Wire rod use risk evaluation method based on inclusion distribution
CN113658092A (en) * 2021-05-13 2021-11-16 湖南莱塞智能装备有限公司 Aluminum electrolytic capacitor defect detection method based on image processing
CN113686890A (en) * 2020-05-18 2021-11-23 泰连服务有限公司 Methods of inspecting parts with visual inspection systems
CN113781430A (en) * 2021-09-09 2021-12-10 北京云屿科技有限公司 Glove surface defect detection method and system based on deep learning
CN114179320A (en) * 2021-11-11 2022-03-15 苏州精思博智人工智能科技有限公司 Automatic adjusting method for technological parameters of injection molding machine in combination with visual detection
CN114663433A (en) * 2022-05-25 2022-06-24 山东科技大学 Method and device for detecting running state of roller cage shoe, computer equipment and medium
CN114943704A (en) * 2022-05-24 2022-08-26 蜂巢能源科技股份有限公司 Method, device and equipment for detecting defects of battery electrode die-cut piece
CN115147327A (en) * 2021-03-29 2022-10-04 株式会社理光 Method, apparatus and storage medium for detecting defects in arcuate edges of objects
CN115147429A (en) * 2022-09-07 2022-10-04 深圳市欣冠精密技术有限公司 Streak detection method for optical glass preform
CN115456956A (en) * 2022-08-19 2022-12-09 浙江华周智能装备有限公司 Method and device for detecting scratches of liquid crystal display and storage medium
CN115760699A (en) * 2022-10-20 2023-03-07 河南科技大学 Method and system for detecting end face chamfering of shaft workpiece
CN115760782A (en) * 2022-11-16 2023-03-07 华南理工大学 A machine vision-based method for identifying offset defects in in-mold labeling
CN115906557A (en) * 2022-10-25 2023-04-04 吉林大学 An Intelligent Detection Method for Determining Forming Defects by Contour Deviation of Stamping Blank
CN116433701A (en) * 2023-06-15 2023-07-14 武汉中观自动化科技有限公司 A method, device, equipment and storage medium for extracting workpiece hole contour
CN116678893A (en) * 2023-04-23 2023-09-01 盐城工学院 A method for detecting surface defects of hot-rolled strip steel based on deep learning
CN116698842A (en) * 2023-03-31 2023-09-05 中国长江电力股份有限公司 System and processing method of hydraulic hoist piston rod rust detection device
CN117173107A (en) * 2023-08-09 2023-12-05 上海移远通信技术股份有限公司 Product defect detection methods, devices, equipment, systems and readable storage media
CN117593302A (en) * 2024-01-18 2024-02-23 新西旺智能科技(深圳)有限公司 A defective parts traceability method and system
CN118887271A (en) * 2024-07-10 2024-11-01 盐城工学院 Flange size measurement method and system based on sub-pixel accuracy of machine vision
CN119919743A (en) * 2025-04-02 2025-05-02 苏州工学院 A general machine vision detection algorithm, device and equipment

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1898555A (en) * 2003-12-22 2007-01-17 玛机统丽公司 Substrate inspection device
CN101334263A (en) * 2008-07-22 2008-12-31 东南大学 Method for locating the center of a circular target
CN104359403A (en) * 2014-11-21 2015-02-18 天津工业大学 Plane part size measurement method based on sub-pixel edge algorithm
CN105701492A (en) * 2014-11-25 2016-06-22 宁波舜宇光电信息有限公司 Machine vision identification system and implementation method thereof
CN105865344A (en) * 2016-06-13 2016-08-17 长春工业大学 Workpiece dimension measuring method and device based on machine vision
CN106952262A (en) * 2017-04-25 2017-07-14 大连理工大学 A Method of Analyzing Ship Plate Machining Accuracy Based on Stereo Vision
US20180293721A1 (en) * 2017-04-07 2018-10-11 Kla-Tencor Corporation Contour based defect detection
CN108830838A (en) * 2018-05-28 2018-11-16 江苏大学 A kind of pcb board incompleteness Trigger jitter detection method of sub-pixel
CN109141232A (en) * 2018-08-07 2019-01-04 常州好迪机械有限公司 A kind of circle plate casting online test method based on machine vision

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1898555A (en) * 2003-12-22 2007-01-17 玛机统丽公司 Substrate inspection device
CN101334263A (en) * 2008-07-22 2008-12-31 东南大学 Method for locating the center of a circular target
CN104359403A (en) * 2014-11-21 2015-02-18 天津工业大学 Plane part size measurement method based on sub-pixel edge algorithm
CN105701492A (en) * 2014-11-25 2016-06-22 宁波舜宇光电信息有限公司 Machine vision identification system and implementation method thereof
CN105865344A (en) * 2016-06-13 2016-08-17 长春工业大学 Workpiece dimension measuring method and device based on machine vision
US20180293721A1 (en) * 2017-04-07 2018-10-11 Kla-Tencor Corporation Contour based defect detection
CN106952262A (en) * 2017-04-25 2017-07-14 大连理工大学 A Method of Analyzing Ship Plate Machining Accuracy Based on Stereo Vision
CN108830838A (en) * 2018-05-28 2018-11-16 江苏大学 A kind of pcb board incompleteness Trigger jitter detection method of sub-pixel
CN109141232A (en) * 2018-08-07 2019-01-04 常州好迪机械有限公司 A kind of circle plate casting online test method based on machine vision

Non-Patent Citations (8)

* Cited by examiner, † Cited by third party
Title
GENG LEI等: "Machine Vision Detection Method for Surface Defects of Automobile Stamping Parts", 《AMERICAN SCIENTIFIC RESEARCH JOURNAL FOR ENGINEERING, TECHNOLOGY, AND SCIENCES》 *
JIAN CHUANXIA等: "Automatic surface defect detection for mobile phone screen glass based on machine vision", 《APPLIED SOFT COMPUTING》 *
李帅等: "基于机器视觉的纽扣缺陷检测算法研究", 《科技创新与应用》 *
李长云等: "《智能感知技术及在电气工程中的应用》", 31 May 2017, 电子科技大学出版社 *
汤勃等: "机器视觉表面缺陷检测综述", 《中国图象图形学报》 *
耿磊等: "上下边缘区分的平面钣金零件尺寸测量方法", 《红外与激光工程》 *
郑树泉等: "《工业智能技术与应用》", 31 January 2019, 上海科学技术出版社 *
郭联金等: "基于亚像素的PCB表面质量检测", 《电子质量》 *

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443791A (en) * 2019-08-02 2019-11-12 西安工程大学 A kind of workpiece inspection method and its detection device based on deep learning network
CN110443791B (en) * 2019-08-02 2023-04-07 西安工程大学 Workpiece detection method and device based on deep learning network
CN110441318A (en) * 2019-08-22 2019-11-12 郑州大学 A kind of chemical fibre spinneret hole defect inspection method based on machine vision
CN110687120A (en) * 2019-09-18 2020-01-14 浙江工商大学 Flange appearance quality inspection system
CN110602462A (en) * 2019-09-27 2019-12-20 南京工程学院 Industrial image detection device and method based on AI
CN110602462B (en) * 2019-09-27 2020-11-03 南京工程学院 An AI-based industrial image detection device and method
CN110579184A (en) * 2019-10-24 2019-12-17 深圳宇骏视觉智能科技有限公司 A product appearance online detection device and its use method
CN111062915A (en) * 2019-12-03 2020-04-24 浙江工业大学 A real-time steel pipe defect detection method based on improved YOLOv3 model
CN111062915B (en) * 2019-12-03 2023-10-24 浙江工业大学 Real-time steel pipe defect detection method based on improved YOLOv3 model
CN111189854A (en) * 2020-04-13 2020-05-22 征图新视(江苏)科技股份有限公司 Defect layering detection method of automatic glass cover plate detection system
CN111189854B (en) * 2020-04-13 2020-08-07 征图新视(江苏)科技股份有限公司 Defect layering detection method of automatic glass cover plate detection system
CN113686890A (en) * 2020-05-18 2021-11-23 泰连服务有限公司 Methods of inspecting parts with visual inspection systems
CN111784662A (en) * 2020-06-29 2020-10-16 深圳至峰精密制造有限公司 Workpiece identification method, device, computer equipment and storage medium
CN112508939A (en) * 2020-12-22 2021-03-16 郑州金惠计算机系统工程有限公司 Flange surface defect detection method, system and equipment
CN112508939B (en) * 2020-12-22 2023-01-20 郑州金惠计算机系统工程有限公司 Flange surface defect detection method, system and equipment
CN112700440B (en) * 2021-01-18 2022-11-04 上海闻泰信息技术有限公司 Object defect detection method, device, computer equipment and storage medium
CN112700440A (en) * 2021-01-18 2021-04-23 上海闻泰信息技术有限公司 Object defect detection method and device, computer equipment and storage medium
CN115147327B (en) * 2021-03-29 2025-06-03 株式会社理光 Method, device and storage medium for detecting defects of arc-shaped edges of objects
CN115147327A (en) * 2021-03-29 2022-10-04 株式会社理光 Method, apparatus and storage medium for detecting defects in arcuate edges of objects
CN113658092A (en) * 2021-05-13 2021-11-16 湖南莱塞智能装备有限公司 Aluminum electrolytic capacitor defect detection method based on image processing
CN113588681A (en) * 2021-07-30 2021-11-02 广东韶钢松山股份有限公司 Wire rod use risk evaluation method based on inclusion distribution
CN113588681B (en) * 2021-07-30 2024-03-12 广东韶钢松山股份有限公司 Wire usage risk evaluation method based on inclusion distribution
CN113781430A (en) * 2021-09-09 2021-12-10 北京云屿科技有限公司 Glove surface defect detection method and system based on deep learning
CN113781430B (en) * 2021-09-09 2023-08-25 北京云屿科技有限公司 Glove surface defect detection method and system based on deep learning
CN114179320A (en) * 2021-11-11 2022-03-15 苏州精思博智人工智能科技有限公司 Automatic adjusting method for technological parameters of injection molding machine in combination with visual detection
CN114943704A (en) * 2022-05-24 2022-08-26 蜂巢能源科技股份有限公司 Method, device and equipment for detecting defects of battery electrode die-cut piece
CN114663433A (en) * 2022-05-25 2022-06-24 山东科技大学 Method and device for detecting running state of roller cage shoe, computer equipment and medium
CN115456956A (en) * 2022-08-19 2022-12-09 浙江华周智能装备有限公司 Method and device for detecting scratches of liquid crystal display and storage medium
CN115456956B (en) * 2022-08-19 2024-05-28 浙江华周智能装备有限公司 Liquid crystal screen scratch detection method, device and storage medium
CN115147429B (en) * 2022-09-07 2022-11-08 深圳市欣冠精密技术有限公司 Streak detection method for optical glass preform
CN115147429A (en) * 2022-09-07 2022-10-04 深圳市欣冠精密技术有限公司 Streak detection method for optical glass preform
CN115760699A (en) * 2022-10-20 2023-03-07 河南科技大学 Method and system for detecting end face chamfering of shaft workpiece
CN115906557B (en) * 2022-10-25 2025-07-11 吉林大学 An intelligent detection method for determining forming defects by contour deviation of stamping blank
CN115906557A (en) * 2022-10-25 2023-04-04 吉林大学 An Intelligent Detection Method for Determining Forming Defects by Contour Deviation of Stamping Blank
CN115760782B (en) * 2022-11-16 2023-06-16 华南理工大学 Machine vision-based in-mold labeling offset defect identification method
CN115760782A (en) * 2022-11-16 2023-03-07 华南理工大学 A machine vision-based method for identifying offset defects in in-mold labeling
CN116698842A (en) * 2023-03-31 2023-09-05 中国长江电力股份有限公司 System and processing method of hydraulic hoist piston rod rust detection device
CN116698842B (en) * 2023-03-31 2024-08-13 中国长江电力股份有限公司 System and processing method of hydraulic hoist piston rod rust detection device
CN116678893A (en) * 2023-04-23 2023-09-01 盐城工学院 A method for detecting surface defects of hot-rolled strip steel based on deep learning
CN116433701B (en) * 2023-06-15 2023-10-10 武汉中观自动化科技有限公司 A method, device, equipment and storage medium for extracting the contour of a workpiece hole
CN116433701A (en) * 2023-06-15 2023-07-14 武汉中观自动化科技有限公司 A method, device, equipment and storage medium for extracting workpiece hole contour
CN117173107A (en) * 2023-08-09 2023-12-05 上海移远通信技术股份有限公司 Product defect detection methods, devices, equipment, systems and readable storage media
CN117593302B (en) * 2024-01-18 2024-05-24 新西旺智能科技(深圳)有限公司 A defective parts tracing method and system
CN117593302A (en) * 2024-01-18 2024-02-23 新西旺智能科技(深圳)有限公司 A defective parts traceability method and system
CN118887271A (en) * 2024-07-10 2024-11-01 盐城工学院 Flange size measurement method and system based on sub-pixel accuracy of machine vision
CN119919743A (en) * 2025-04-02 2025-05-02 苏州工学院 A general machine vision detection algorithm, device and equipment
CN119919743B (en) * 2025-04-02 2025-06-20 苏州工学院 Machine vision general detection algorithm, device and equipment

Similar Documents

Publication Publication Date Title
CN109900711A (en) Workpiece, defect detection method based on machine vision
CN115294113B (en) A method for quality inspection of wood veneer panels
CN105067638B (en) Tire fetal membrane face character defect inspection method based on machine vision
CN113340909B (en) Glue line defect detection method based on machine vision
CN117649404A (en) A quality inspection method and system for pharmaceutical packaging boxes based on image data analysis
CN102095733B (en) Machine vision-based intelligent detection method for surface defect of bottle cap
CN111862037A (en) Method and system for geometric feature detection of precision hole parts based on machine vision
CN103471910B (en) A kind of elongation at break of metal material intelligent test method followed the tracks of based on random point
CN105335973B (en) Apply to the visual processing method of strip machining production line
CN108844961A (en) A kind of temperature controller case vision detection system and method
CN107239742B (en) Method for calculating scale value of instrument pointer
CN107705293A (en) A kind of hardware dimension measurement method based on CCD area array cameras vision-based detections
CN107392929B (en) An intelligent target detection and size measurement method based on human visual model
CN116758067B (en) Metal structural member detection method based on feature matching
CN108846397A (en) A kind of cable semi-conductive layer automatic testing method based on image procossing
CN114627080B (en) Defect detection method of vehicle stamping parts based on computer vision
CN115599844B (en) A Visual Detection Method for Misfitting or Missing Aircraft Wing Connectors
CN119919410B (en) Method for measuring size and appearance of O-rings
CN115375608B (en) Detection method and device, detection equipment and storage medium
CN108520514A (en) Printed circuit board electronics member device consistency detecting method based on computer vision
CN116563279B (en) Measuring switch detection method based on computer vision
CN116091506B (en) Machine vision defect quality inspection method based on YOLOV5
CN112819844A (en) Image edge detection method and device
CN120707538A (en) Multimodal dynamic defect detection system for industrial products based on edge computing
CN113808104A (en) Block-based metal surface defect detection method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20190618