Monocular camera 3D detection method and system based on vision algorithm

Abstract

The invention discloses a monocular camera 3D detection method and system based on a vision algorithm, and relates to the technical field of vision detection.The monocular camera 3D detection system comprises a data collection module, a space calculation module, a feature analysis module and a 3D detection module, the data collection module is used for collecting a three-dimensional model library and obtaining acceleration changes of a monocular camera on three axes in real time; and the space calculation module is used for calculating the relative space position distribution of the monocular camera when each image is shot, and identifying a missing area of the mapping model by combining modeling software. The method has the beneficial effects that the key features in the 2D image are accurately extracted by calculating the relative spatial position distribution of the monocular camera when each image is shot, identifying the missing region of the mapping model in combination with modeling software, establishing the multivariate linear relationship of the key features by using the three-dimensional model library, and calculating and filling the information of the missing region, so that the accuracy of the key features in the 2D image is improved. The precision of the three-dimensional model of the automobile hub is improved, and the accuracy of 3D detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of visual detection technology, and in particular to a monocular camera 3D detection method and system based on a visual algorithm. Background Art

[0002] In the important field of industrial production and quality inspection, accurate 3D inspection of automobile wheels is of great significance. The existing technology relies on laser scanners to achieve 3D inspection of automobile wheels. However, this method has some disadvantages, such as expensive equipment and high purchase cost. On the other hand, in actual use, the operation process is relatively complicated and has high requirements for the use environment. These factors limit its wider application to a certain extent.

[0003] At present, when testing automobile wheels, most of them use manual testing or simpler testing methods based on 2D images. Manual testing is inefficient, and subjective factors such as the experience and status of the tester will have a great impact on the test results, making it difficult to ensure the stability and accuracy of the test results. The 2D image-based testing method cannot obtain the relative spatial position distribution of the captured image, making it difficult to construct a complete and accurate wheel status, so it is difficult to achieve a comprehensive and accurate test effect.

[0004] Nowadays, with the continuous development of computer vision technology, by converting the 2D image information of the camera into a 3D spatial representation, a 3D model of the car wheel (such as the car wheel) is constructed to complete a series of 3D detection tasks such as the size measurement and defect detection of the car wheel. Therefore, how to accurately obtain the key features of the car wheel from the 2D image is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of this section is to provide a monocular camera 3D detection method and system based on visual algorithms, which can accurately obtain the key features of automobile wheels from 2D images.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a monocular camera 3D detection method based on a visual algorithm, comprising the following steps: S100, collecting a three-dimensional model library, using a monocular camera to shoot a car wheel hub from multiple perspectives, and obtaining the acceleration changes of the monocular camera on three axes in real time; S200, calculating the relative spatial position distribution of the monocular camera when shooting each image, and identifying the missing area of ​​the mapping model in combination with the modeling software; S300, extracting the key features of the car wheel hub from the image through a visual algorithm, and establishing a multivariate linear relationship in combination with the key features of similar car wheel hub models in the three-dimensional model library; S400, filling in the information of the missing area according to the multivariate linear relationship, establishing a three-dimensional model of the car wheel hub, and performing 3D detection on the car wheel hub.

[0007] As a preferred solution of the monocular camera 3D detection method based on the vision algorithm according to the present invention, in S200, the specific steps are as follows: S201. All the images are arranged in the order of shooting time, the changes in the triaxial acceleration of the monocular camera within the shooting time periods corresponding to adjacent images are retrieved, the spatial position of the monocular camera when shooting the first image is set as the origin, and the spatial positions of each image relative to the origin when shooting are analyzed one by one; S202. Use modeling software to establish a mapping model of the vehicle wheel hub in combination with the content and spatial position distribution of all the images, and mark the missing areas in the mapping model as abnormal areas.

[0008] As a preferred solution of the monocular camera 3D detection method based on the vision algorithm according to the present invention, in S300, the specific steps are as follows: S301. Obtain the key features of the vehicle wheel hub, retrieve the 3D model with the highest similarity in the 3D model library according to the key features, map the abnormal areas onto the 3D model, analyze the feature relationships between the normal areas and the abnormal areas in the 3D model as the training set, and establish a multiple linear relationship formula: Y n = aY 1 + bY 2 + C. In the formula, Y n represents the key feature value of the abnormal area, Y 1 and Y 2 respectively represent the key feature values corresponding to two different normal areas, a and b are regression coefficients, and C is a constant; S302. Substitute the key features of each normal area in the vehicle wheel hub mapping model into the above formula to obtain the key feature value of the abnormal area; S303. Use the key feature value of the abnormal area to fill the missing areas in the vehicle wheel hub mapping model.

[0009] As a preferred solution of the monocular camera 3D detection method based on the vision algorithm according to the present invention, in S300, the key features include the size features, shape features of the vehicle wheel hub, and the corresponding center points, corner points, and contour lines of the features.

[0010] As a preferred solution of the monocular camera 3D detection method based on the vision algorithm according to the present invention, the size features of the vehicle wheel hub include the hub diameter, the width and thickness of the spokes, the size and spacing of the mounting holes, and the depth of the mounting holes.

[0011] As a preferred solution of the monocular camera 3D detection system based on the vision algorithm according to the present invention, it includes a data acquisition module, a spatial calculation module, a feature analysis module, and a 3D detection module.

[0012] As a preferred solution of the monocular camera 3D detection system based on the vision algorithm of the present invention, wherein: the data acquisition module is used to acquire a three-dimensional model library and obtain the acceleration changes of the monocular camera on the three axes in real time.

[0013] As a preferred solution of the monocular camera 3D detection system based on the vision algorithm of the present invention, wherein: the space calculation module is used to calculate the relative spatial position distribution of the monocular camera when taking each image, and combine with the modeling software to identify the missing areas of the mapping model.

[0014] As a preferred solution of the monocular camera 3D detection system based on the vision algorithm of the present invention, wherein: the feature analysis module extracts the key features of the car wheel hub from the image through the vision algorithm, and establishes a multiple linear relationship by combining the key features of the similar car wheel hub models in the three-dimensional model library.

[0015] As a preferred solution of the monocular camera 3D detection system based on the vision algorithm of the present invention, wherein: the 3D detection module fills in the information of the missing areas according to the multiple linear relationship, establishes a three-dimensional model of the car wheel hub, and performs 3D detection on the car wheel hub.

[0016] Advantages of the present invention:

[0017] 1. By calculating the relative spatial position distribution of the monocular camera when taking each image, combining with the modeling software to identify the missing areas of the mapping model, using the three-dimensional model library, establishing a multiple linear relationship of the key features, calculating and filling in the information of the missing areas, the key features in the 2D image are accurately extracted, the accuracy of the three-dimensional model of the car wheel hub is improved, and the accuracy of the 3D detection is enhanced.

[0018] 2. By using the monocular camera as the image acquisition device, the detection cost is low, the operation is simple, the requirements for the use environment are relatively low, and the versatility is high. It is not only applicable to the detection of car wheel hubs, but also applicable to the detection of other objects such as mechanical parts and electronic products. Brief Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. The following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative and laborious efforts. Among them:

[0020] Figure 1 It is a schematic flow chart of the monocular camera 3D detection method based on the vision algorithm of the present invention. Detailed Embodiments

[0021] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be provided in conjunction with the accompanying drawings of the specification.

[0022] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. The "embodiments" referred to herein mean specific features, structures, or characteristics that may be included in at least one implementation of the present invention.

[0023] Embodiment 1

[0024] Referring to Figure 1 , this embodiment provides a monocular camera 3D detection method based on a vision algorithm, which specifically includes the following steps: S100, collect a three-dimensional model library, use a monocular camera to take multi-angle photos of an automotive wheel hub, and obtain the acceleration changes of the monocular camera on the three axes in real time; S200, calculate the relative spatial position distribution of the monocular camera when taking each image, and combine a modeling software to identify the missing areas of the mapping model; S300, extract the key features of the automotive wheel hub from the image through a vision algorithm, and establish a multiple linear relationship in combination with the key features of similar automotive wheel hub models in the three-dimensional model library; S400, fill in the information of the missing areas according to the multiple linear relationship, establish a three-dimensional model of the automotive wheel hub, and perform 3D detection on the automotive wheel hub.

[0025] In S200, the specific steps are as follows: S201, arrange all the images in the order of shooting time, retrieve the acceleration changes of the three axes of the monocular camera during the shooting time periods corresponding to adjacent images, set the spatial position of the monocular camera when taking the first image as the origin, and analyze the spatial positions of each image relative to the origin when shooting one by one; S202, use a modeling software to establish a mapping model of the automotive wheel hub in combination with the content and spatial position distribution of all the images, and mark the missing areas in the mapping model as abnormal areas.

[0026] In S300, the specific steps are as follows: S301, obtain the key features of the automotive wheel hub, retrieve the three-dimensional model with the highest similarity in the three-dimensional model library according to the key features, map the abnormal areas to this three-dimensional model, analyze the feature relationship between the normal areas and the abnormal areas in the three-dimensional model, use it as a training set, and establish a multiple linear relationship formula: Y n = aY 1 + bY 2 + C. In the formula, Y n represents the key feature value of the abnormal area, Y 1 and Y 2respectively represent the key feature values corresponding to two different normal regions, a and b are regression coefficients, and C is a constant; S302. Substitute the key features of each normal region in the automobile wheel hub mapping model into the above formula to obtain the key feature values of the abnormal region; S303. Use the key feature values of the abnormal region to fill the missing regions in the automobile wheel hub mapping model.

[0027] In S300, the key features include the size features, shape features of the automobile wheel hub, and the corresponding center points, corner points, and contour lines of the features.

[0028] The size features of the automobile wheel hub include the wheel hub diameter, the width and thickness of the spokes, the size and spacing of the mounting holes, and the depth of the mounting holes.

[0029] The shape features of the automobile wheel hub include roundness, flatness, and contour. The overall roundness of the wheel hub reflects the coincidence degree between its rotation center and geometric center. Flatness refers to the flatness of the mounting surface, braking surface, etc. of the wheel hub. Contour refers to the variation of the measured actual contour relative to the ideal contour.

[0030] Embodiment 2

[0031] This embodiment provides a monocular camera 3D detection system based on a vision algorithm, which specifically includes a data acquisition module, a space calculation module, a feature analysis module, and a 3D detection module.

[0032] The data acquisition module is used to collect a three-dimensional model library and obtain the acceleration changes of the monocular camera on the three axes in real time.

[0033] The space calculation module is used to calculate the relative spatial position distribution of the monocular camera when taking each image, and identify the missing regions of the mapping model in combination with the modeling software.

[0034] The feature analysis module extracts the key features of the automobile wheel hub from the image through a vision algorithm, and establishes a multiple linear relationship in combination with the key features of similar automobile wheel hub models in the three-dimensional model library.

[0035] The 3D detection module fills the information of the missing regions according to the multiple linear relationship, establishes a three-dimensional model of the automobile wheel hub, and performs 3D detection on the automobile wheel hub.

[0036] Use a monocular camera to collect images of the automobile wheel hub from multiple angles, and perform image preprocessing to enhance the image quality. Through a vision algorithm, extract features from the image, identify the key features of the automobile wheel hub, such as the contour of the automobile wheel hub, the center points of the hole positions, the corner points, and the corresponding depth information. Construct a 3D model of the automobile wheel hub based on the extracted features, convert the 2D image information into a 3D space representation, realize 3D detection tasks such as size measurement and defect detection of the automobile wheel hub, and output the detection results.

[0037] Importantly, although only a few embodiments are described in detail in this disclosure, those skilled in the art who refer to this disclosure should readily understand that many modifications are possible without materially departing from the subject matter described in this application. For example, the dimensions, structures, shapes, and proportions of various components, as well as temperature, pressure, installation arrangements, use of materials, colors, orientation changes, etc.; for example, components shown as integrally formed can be composed of multiple parts or components, and the positions of the components can be inverted or otherwise changed; accordingly, all such modifications should be included within the scope of the present invention, and other substitutions, modifications, changes, and omissions can be made in the design, operating conditions, and arrangements of the exemplary embodiments without departing from the scope of the present invention.

Claims

1. A monocular camera 3D detection method based on a visual algorithm, characterized in that: The following steps are involved: S100, collect the 3D model library, use the monocular camera to shoot the car wheel from multiple perspectives, and obtain the acceleration changes of the monocular camera on three axes in real time; S200, calculating the relative spatial position distribution of the monocular camera when taking each image, and identifying the missing area of ​​the mapping model in combination with the modeling software; S300, extracting key features of the automobile wheel from the image by using a visual algorithm, and establishing a multivariate linear relationship by combining key features of similar automobile wheel models in a three-dimensional model library; S400, filling in the missing area information according to the multivariate linear relationship, establishing a three-dimensional model of the automobile wheel hub, and performing 3D detection on the automobile wheel hub.

2. The monocular camera 3D detection method based on visual algorithm as claimed in claim 1, characterized in that: In S200, the specific steps are as follows: S201, all images are arranged in order of shooting time, the three-axis acceleration changes of the monocular camera in the corresponding shooting time period of adjacent images are retrieved, the spatial position of the monocular camera when the first image is shot is set as the origin, and the spatial position of each image relative to the origin when it is shot is analyzed one by one; S202, using modeling software to combine the content and spatial position distribution of all images to establish a mapping model of the automobile wheel hub, and marking the missing areas in the mapping model as abnormal areas.

3. The monocular camera 3D detection method based on visual algorithm as claimed in claim 2, characterized in that: In S300, the specific steps are as follows: S301, obtain key features of the automobile wheel hub, retrieve the three-dimensional model with the highest similarity in the three-dimensional model library according to the key features, map the abnormal area to the three-dimensional model, analyze the feature relationship between the normal area and the abnormal area in the three-dimensional model as a training set, and establish a multivariate linear relationship formula: Y n =aY1+bY2+C Where Y n represents the key eigenvalue of the abnormal area, Y1 and Y2 represent the key eigenvalues ​​corresponding to two different normal areas, a and b are regression coefficients, and C is a constant; S302, substituting the key features of each normal area in the automobile wheel hub mapping model into the above formula to obtain the key feature value of the abnormal area; S303: Use the key feature values ​​of the abnormal area to fill the missing area in the automobile wheel hub mapping model.

4. The monocular camera 3D detection method based on visual algorithm as claimed in claim 1, characterized in that: In S300, the key features include the size features, shape features, and the center points, corner points, and contour lines corresponding to the features of the automobile wheel hub.

5. The monocular camera 3D detection method based on visual algorithm as claimed in claim 4, characterized in that: The dimensional characteristics of automotive wheels include the hub diameter, spoke width and thickness, mounting hole size and spacing, and mounting hole depth.

6. A monocular camera 3D detection system based on visual algorithms, characterized in that: It includes data acquisition module, spatial calculation module, feature analysis module and 3D detection module.

7. The monocular camera 3D detection system based on visual algorithm as claimed in claim 6, characterized in that: The data acquisition module is used to collect the three-dimensional model library and obtain the acceleration changes of the monocular camera on three axes in real time.

8. The monocular camera 3D detection system based on visual algorithm as claimed in claim 6, characterized in that: The spatial calculation module is used to calculate the relative spatial position distribution of the monocular camera when taking each image, and identify the missing areas of the mapping model in combination with the modeling software.

9. The monocular camera 3D detection system based on visual algorithm as claimed in claim 6, characterized in that: The feature analysis module extracts key features of the automobile wheel hub from the image through a visual algorithm, and establishes a multivariate linear relationship by combining key features of similar automobile wheel hub models in a three-dimensional model library.

10. The monocular camera 3D detection system based on visual algorithm according to claim 6, characterized in that: The 3D detection module fills in the information of the missing area according to the multivariate linear relationship, establishes a three-dimensional model of the automobile wheel hub, and performs 3D detection on the automobile wheel hub.

Images