CN116740615A

CN116740615A - A method for detecting small targets on the apron integrating depth information

Info

Publication number: CN116740615A
Application number: CN202310805592.3A
Authority: CN
Inventors: 樊治国; 夏克江
Original assignee: Qingdao Gaozhong Information Technology Co ltd
Current assignee: Qingdao Gaozhong Information Technology Co ltd
Priority date: 2023-07-03
Filing date: 2023-07-03
Publication date: 2023-09-12

Abstract

The invention discloses a method for detecting a small target of an apron by fusing depth information, which specifically comprises the following steps: s1, measuring small target sizes such as a wheel guard, a reflecting cone and the like in advance; s2, shooting left and right images by using a binocular camera, and acquiring a parallax image of a current scene through a binocular stereo matching algorithm; s3, calculating a corresponding U-V parallax map according to the parallax map, and acquiring small targets such as a wheel guard, a reflecting cone and the like on the apron on the U-V parallax map; s4, extracting a target image on the corresponding RGB image according to the parallax image small target detection result; s5, inputting the target image into a trained image classification model to obtain a target class result; according to the invention, the small target area is extracted through binocular stereoscopic vision, so that the problem of detection of small targets such as a wheel guard, a reflecting cone and the like is effectively solved; by training a simple classifier network, the input small target images are classified, and category information is output.

Description

Method for detecting small targets of apron by fusing depth information

Technical Field

The invention relates to the technical field of target detection methods, in particular to a method for detecting a small target of an apron by fusing depth information.

Background

The patent of automatic identification method for placing a reflecting cone on an air park (application number CN201811228974. X) predicts the standard position of the reflecting cone on a picture to be detected by using an image registration method based on the determination that the spatial relationship between the reflecting cone and the relevant part of an airplane is relatively fixed. The method has obvious defects: in a monitoring image, a reflecting cone belongs to a small target, and characteristic points of the small target are difficult to extract by an image registration method, so that accuracy of the method is difficult to guarantee.

The patent "automatic identification method of the wheel gear arrangement specification on the civil aircraft parking apron" (application number CN 202110295767.1) discloses an automatic identification method of the wheel gear arrangement specification on the civil aircraft parking apron, which requires that the wheel gear is manufactured to have a specified reflective mark and color, then a deep learning model is adopted to detect the position of the wheel gear, and whether the wheel gear exists or not is detected at the position of the wheel gear. The method has obvious defects: firstly, the method requires that the wheel guard is made to have specified reflective marks and colors, so that the workload of airport equipment maintenance personnel and the airport operation cost are increased; secondly, the target detection technology based on the neural network is easily affected by weather, illumination, target size and the like, and the detection accuracy is difficult to guarantee.

Disclosure of Invention

The invention aims to solve the technical problem of providing an apron small target detection method integrating depth information, which solves the problems of accuracy and recall rate of the existing target detection algorithm in solving the problem of small target detection.

The method for detecting the small target of the apron by fusing depth information is realized by the following technical scheme: the method specifically comprises the following steps:

s1, measuring small target sizes such as a wheel guard, a reflecting cone and the like in advance;

s2, shooting left and right images by using a binocular camera, and acquiring a parallax image of a current scene through a binocular stereo matching algorithm;

s3, calculating a corresponding U-V parallax map according to the parallax map, and acquiring small targets such as a wheel guard, a reflecting cone and the like on the apron on the U-V parallax map;

s4, extracting a target image on the corresponding RGB image according to the parallax image small target detection result;

s5, inputting the target image into a trained image classification model to obtain a target class result.

As an preferable technical scheme, on the premise of the step S2, a dense parallax map of the apron area is obtained by a binocular stereoscopic vision technology.

Step S3, extracting the position of the parking apron pavement from the V disparity map; detecting all targets of the tarmac in the U-disparity map.

As a preferable technical scheme, step S4, according to the target position detected in the UV parallax map, combines with a preset small target size standard, filters all targets to obtain small target information meeting the preset small target information, acquires a corresponding position screenshot in the RGB image, and stores the small target screenshot.

As a preferable technical scheme, step S5 is to send the small target screenshot into a corresponding image classification network, output target class information and obtain a final small target detection result.

The beneficial effects of the invention are as follows:

1. the small target area is extracted through binocular stereoscopic vision, so that the problem of detection of small targets such as a wheel guard, a reflecting cone and the like is effectively solved;

2. further, by training a simple classifier network, classifying the input small target image and outputting class information;

3. the detection method of the apron wheel block and the reflection cone which are integrated with the depth information greatly improves the detection rate and the accuracy of small targets of the apron. Meanwhile, as the stereoscopic vision technology is introduced to extract the small target area, the operation efficiency of the algorithm is improved, and the training complexity of the algorithm model is reduced.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a method for detecting a small target of an apron by fusing depth information;

FIG. 2 is an original diagram of a method for detecting a small target of an apron by fusing depth information;

fig. 3 is a parallax diagram of the method for detecting a small target of an apron by fusing depth information;

fig. 4 is a UV parallax diagram of the tarmac small target detection method of the present invention incorporating depth information;

FIG. 5 is a schematic diagram of the invention for extracting small objects in a UV disparity map;

FIG. 6 is a small target image location;

fig. 7 is a small object class.

Detailed Description

All of the features disclosed in this specification, or all of the steps in a method or process disclosed, may be combined in any combination, except for mutually exclusive features and/or steps.

Any feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. That is, each feature is one example only of a generic series of equivalent or similar features, unless expressly stated otherwise.

In the description of the present invention, it should be understood that the terms "one end," "the other end," "the outer side," "the upper," "the inner side," "the horizontal," "coaxial," "the center," "the end," "the length," "the outer end," and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, merely to facilitate description of the present invention and simplify the description, and do not indicate or imply that the device or element being referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention.

Furthermore, in the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.

Terms such as "upper," "lower," and the like used herein to refer to a spatially relative position are used for ease of description to describe one element or feature's relationship to another element or feature as illustrated in the figures. The term spatially relative position may be intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if the device in the figures is turned over, elements described as "below" or "beneath" other elements or features would then be oriented "above" the other elements or features. Thus, the exemplary term "below" can encompass both an orientation of above and below. The device may be otherwise oriented (rotated 90 degrees or other orientations) and the spatially relative descriptors used herein interpreted accordingly.

In the present invention, unless explicitly specified and limited otherwise, the terms "disposed," "coupled," "connected," "plugged," and the like are to be construed broadly and may be, for example, fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly, through intermediaries, or both, may be in communication with each other or in interaction with each other, unless expressly defined otherwise. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.

As shown in fig. 1 to 7, the method for detecting the small target of the apron by fusing depth information specifically comprises the following steps:

In this embodiment, on the premise of step S2, a dense parallax map of the apron area is obtained by binocular stereoscopic vision technology.

In the embodiment, step S3, extracting the position of the parking apron pavement from the V parallax map; and combining the extracted tarmac pavement information, and acquiring all targets on the tarmac in the U parallax map. The large-scale light-emitting device comprises small targets such as a wheel guard, a light reflecting cone, an operator and the like, and large targets such as an airplane, a special vehicle and a gallery bridge and the like.

As shown in fig. 4 and 5, in the embodiment, step S4, according to the target position detected in the UV parallax map, a preset small target size standard is combined, small target information meeting the preset is obtained by filtering in all targets, a corresponding position screenshot is obtained in the RGB image, and the small target screenshot is saved;

and sending the small target screenshot to a corresponding image classification network, and outputting target class information to obtain a final small target detection result.

The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any changes or substitutions that do not undergo the inventive effort should be construed as falling within the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope defined by the claims.

Claims

1. A method for detecting small targets on an apron that integrates depth information, which is characterized by: specifically including the following steps:

S1. Pre-measure the size of small targets such as wheel chocks and reflective cones;

S2. Use the binocular camera to capture the left and right images, and obtain the disparity map of the current scene through the binocular stereo matching algorithm;

S3. Based on the disparity map, calculate the corresponding U-V disparity map, and obtain small targets such as wheel blocks and reflective cones on the apron on the U-V disparity map;

S4. Based on the small target detection results of the disparity map, extract the target image on the corresponding RGB image;

S5. Input the target image into the trained image classification model to obtain the target category result.

2. The method for detecting small targets on an apron by integrating depth information according to claim 1, characterized in that, under the premise of step S2, a dense disparity map of the apron area is obtained through binocular stereo vision technology.

3. The method for detecting small targets on an apron by integrating depth information according to claim 1, characterized in that: in step S3, the location of the apron road surface is extracted in the V disparity map; and all the locations on the apron are detected in the U disparity map. Target.

4. The airport small target detection method integrating depth information according to claim 1, characterized in that: the step S4 is based on the target position detected in the UV disparity map, combined with the preset small target size. Standard, filter all targets to obtain small target information that meets the preset, obtain corresponding position screenshots in the RGB image, and save the small target screenshots.

5. The airport small target detection method integrating depth information according to claim 1, characterized in that: in step S5, the small target screenshot is sent to the corresponding image classification network, and the target category information is output to obtain the final Small target detection results.