CN111553268A - Vehicle component identification method, device, computer equipment and storage medium - Google Patents

Abstract

The application relates to the field of artificial intelligence, is applied to the field of intelligent traffic, and provides a vehicle component identification method and device based on image detection, computer equipment and a storage medium. The method comprises the following steps: and identifying the target vehicle picture to be detected to obtain an identification result, determining the display area and the vehicle model of the target vehicle according to the identification result, and determining the shooting angle of the target vehicle picture to be detected according to the identification result and the vehicle model. And acquiring the standard relative position relation and the standard contour line of each key component according to the vehicle model, and correcting the display area according to the standard relative position relation and the shooting angle. And performing edge approximation comparison on the contour line of the target vehicle in the corrected display area and the standard contour line to determine the actual positions of all key components of the target vehicle. By adopting the method, the actual positions of all key parts of the target vehicle are further determined through edge approximate comparison, and the accuracy of vehicle key part identification is improved.

Description

Vehicle component identification method, device, computer equipment and storage medium

technical field

The present application relates to the technical field of artificial intelligence, and in particular, to a vehicle component identification method, device, computer equipment and storage medium.

Background technique

With the increasing development of the social economy, vehicles are widely used in daily life. If the vehicle is damaged due to natural disasters or accidents during use, the vehicle needs to be assessed for damage to facilitate subsequent claims settlement.

In the vehicle damage assessment scenario, the vehicle exterior parts need to be identified. The traditional identification method is to identify the various parts of the car on the photo of the car damage by receiving the photos of the car damage taken on the spot, and determine the damage of each part of the vehicle appearance. However, due to the complexity of the accident scene and the limitation of shooting technology, it is easy to be unable to accurately determine the location of each part of the damaged vehicle and the damage to the damaged parts.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a vehicle component identification method, device, computer equipment and storage medium that can improve the identification accuracy of vehicle exterior components in a vehicle damage assessment scenario.

A vehicle component identification method, the method comprising:

Obtaining a picture of the target vehicle to be detected, identifying the picture of the target vehicle, and obtaining the identification result of each key component of the picture of the target vehicle;

Determine the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle according to the recognition result;

Determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model;

According to the vehicle model, obtain the standard relative positional relationship and standard contour line of each key component of the corresponding vehicle;

Correcting the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture;

extracting the contour line of the target vehicle in the display area after correction;

The contour line of the target vehicle is compared with the standard contour line by edge approximation to determine the actual position of each key component of the target vehicle.

In one embodiment, a picture of the target vehicle to be detected is obtained, the picture of the target vehicle is identified, and the identification result of each key component of the picture of the target vehicle is obtained, including:

Obtain a picture of the target vehicle to be detected;

Obtain the convolutional neural network model trained by the sample set of vehicle images;

Input the image of the target vehicle into the trained convolutional neural network model, and identify the image of the target vehicle;

Obtain the identification results of each key component of the target vehicle picture;

Also includes:

Upload the identification result to the blockchain.

In one embodiment, the determining, according to the recognition result, the display area of the target vehicle on the picture of the target vehicle, and the vehicle model of the target vehicle, include:

Based on the identification result, determine the relative positions of the key components of the target vehicle;

Determine the display area of the target vehicle on the picture of the target vehicle according to the relative positions of the key components of the target vehicle;

The target vehicle in the display area is extracted, and the vehicle model of the target vehicle is determined.

In one embodiment, the determining the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model includes:

According to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the eigenvector matrix corresponding to each key component of the target vehicle is determined;

According to the vehicle model of the target vehicle, obtain the baseline feature vector of each key component of the corresponding vehicle;

The baseline feature vector is compared with the feature vector matrix, the baseline feature vector is rotated, and the shooting angle of the image of the target vehicle to be detected is determined.

In one of the embodiments, performing an edge approximation comparison between the contour line of the target vehicle and the standard contour line to determine the actual position of each key component of the target vehicle, including:

Performing an approximate edge comparison between the contour line of the target vehicle and the standard contour line to obtain an approximate edge comparison result;

establishing an association relationship between the key components of the target vehicle and the key components of vehicles of the same vehicle model based on the edge approximation comparison result;

According to the association relationship and the standard relative position relationship, the actual position of each key component of the target vehicle is determined.

In one embodiment, the key point angle vector of each key component of the target vehicle is obtained according to the identification result of each key component of the target vehicle, and the feature vector corresponding to each key component of the target vehicle is determined matrix, including:

According to the identification result of each key component of the target vehicle, determine the key point position of each key component;

Extracting key points at the positions of each of the key points, and calculating a key point angle vector between any two key points based on a preset arrangement order;

According to the key point angle vector, determine the relevant angle vector corresponding to the key point;

According to the key point angle vector and the corresponding relevant angle vector, a feature vector matrix corresponding to each key component of the target vehicle is obtained.

In one embodiment, comparing the baseline feature vector with the feature vector matrix, rotating the baseline feature vector, and determining the shooting angle of the image of the target vehicle to be detected includes:

extracting horizontal eigenvectors and vertical eigenvectors of the eigenvector matrix;

Comparing the horizontal feature vector and the vertical feature vector with the baseline feature vector to obtain a comparison result;

Based on the comparison result, determine the rotation angle of the baseline feature vector;

Rotate the baseline feature vector according to the rotation angle, and calculate the geometric mean of the angle vector of the common key point in the process of rotating the baseline vector; the common key point is the common feature vector matrix and the baseline feature vector. key point;

When the geometric mean of the angle vectors of the common key points reaches a preset threshold, the shooting angle of the picture of the target vehicle to be detected is obtained.

A vehicle component identification device comprising:

A recognition result generation module, configured to obtain a picture of the target vehicle to be detected, identify the picture of the target vehicle, and obtain the recognition results of each key component of the picture of the target vehicle;

a display area determination module, configured to determine the display area of the target vehicle on the picture of the target vehicle and the vehicle model of the target vehicle according to the recognition result;

a shooting angle determination module, configured to determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model;

a first acquisition module, configured to acquire, according to the vehicle model, standard relative positional relationships and standard contour lines of each key component of the corresponding vehicle;

a display area correction module, configured to correct the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture;

a target vehicle contour line extraction module, used for extracting the corrected contour line of the target vehicle in the display area;

The actual position determination module of key components is used to compare the contour line of the target vehicle with the standard contour line by approximating the edge to determine the actual position of each key component of the target vehicle.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

Obtaining a picture of the target vehicle to be detected, identifying the picture of the target vehicle, and obtaining the identification result of each key component of the picture of the target vehicle;

Determine the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle according to the recognition result;

Determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model;

According to the vehicle model, obtain the standard relative positional relationship and standard contour line of each key component of the corresponding vehicle;

Correcting the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture;

extracting the contour line of the target vehicle in the display area after correction;

The contour line of the target vehicle is compared with the standard contour line by edge approximation to determine the actual position of each key component of the target vehicle.

A computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

Obtaining a picture of the target vehicle to be detected, identifying the picture of the target vehicle, and obtaining the identification result of each key component of the picture of the target vehicle;

Determine the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle according to the recognition result;

Determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model;

According to the vehicle model, obtain the standard relative positional relationship and standard contour line of each key component of the corresponding vehicle;

Correcting the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture;

extracting the contour line of the target vehicle in the display area after correction;

The contour line of the target vehicle is compared with the standard contour line by edge approximation to determine the actual position of each key component of the target vehicle.

The above-mentioned vehicle component identification method, device, computer equipment and storage medium, by identifying the acquired target vehicle picture to be detected, the identification result of each key component of the target vehicle picture is obtained, and the identification result is uploaded to the blockchain, according to The recognition result determines the display area of the target vehicle on the image of the target vehicle, and the vehicle model of the target vehicle. According to the recognition result and the vehicle model, the shooting angle of the image of the target vehicle to be detected is determined, and according to the vehicle model, the standard relative position relationship and standard contour line of each key component of the corresponding vehicle are obtained. Then, according to the standard relative position relationship and the shooting angle of the target vehicle picture, the display area is corrected to obtain a display area covering all key components of the vehicle as much as possible. By extracting the contour line of the target vehicle in the corrected display area, and performing an approximate edge comparison between the contour line of the target vehicle and the standard contour line, the relationship between the various components can be established. Based on the contour line and the relationship between the components, the actual position of each key component of the target vehicle is further determined, and the accuracy of the identification of the key components of the vehicle is improved.

Description of drawings

1 is an application scenario diagram of a vehicle component identification method in one embodiment;

FIG. 2 is a schematic flowchart of a vehicle component identification method in one embodiment;

3 is a schematic flowchart of obtaining the identification results of each key component of the target vehicle picture in one embodiment;

FIG. 4 is a schematic diagram of marking the key components of the first part of a standard vehicle in one embodiment;

FIG. 5 is a schematic diagram of marking key components of the second part of a standard vehicle in one embodiment;

6 is a schematic flowchart of determining a display area on a picture of a target vehicle and a vehicle model of the target vehicle in one embodiment;

7 is a schematic flowchart of determining the shooting angle of the target vehicle picture in one embodiment;

FIG. 8 is a schematic diagram of the position of key points of key components of the target vehicle in one embodiment;

9 is a schematic diagram of a key point angle vector of a key component of a target vehicle in one embodiment;

10 is a schematic diagram of the baseline feature vector of a sample vehicle in one implementation;

11 is a schematic diagram of a vector comparison of a vehicle component identification method in one embodiment;

12 is a structural block diagram of a vehicle component identification device in one embodiment;

Figure 13 is a diagram of the internal structure of a computer device in one embodiment.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

The vehicle component identification method provided in this application can be applied to the application environment shown in FIG. 1 . The terminal 102 communicates with the server 104 through the network. The server 104 receives the image of the target vehicle to be detected sent by the terminal 102, and identifies the image of the target vehicle to obtain the identification result of each key component of the image of the target vehicle. According to the identification result, the display area of the target vehicle on the image of the target vehicle can be determined, and the vehicle model of the target vehicle, and then the server 104 can determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model, and obtain the standard relative position relationship and standard contour line of each key component of the corresponding vehicle according to the vehicle model, and Correct the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture. By extracting the contour line of the target vehicle in the corrected display area, and comparing the contour line of the target vehicle with the standard contour line, the actual position of each key component of the target vehicle is determined. The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

In one embodiment, as shown in FIG. 2 , a method for identifying vehicle components is provided, which is described by taking the method applied to the server in FIG. 1 as an example, including the following steps:

In step S202, a picture of the target vehicle to be detected is acquired, the picture of the target vehicle is identified, and the identification result of each key component of the picture of the target vehicle is obtained.

Specifically, by acquiring the image of the target vehicle to be detected, and acquiring the convolutional neural network model trained by the vehicle image in the sample set. Then, input the target vehicle image into the trained convolutional neural network model, identify the target vehicle image, and obtain the recognition results of key components of the target vehicle image output by the trained convolutional neural network model.

Further, the convolutional neural network model is trained by using the vehicle pictures of the sample set with each key component marked, and the trained convolutional neural network model can be obtained. Among them, the trained convolutional neural network model can be used to identify the image of the target vehicle, and obtain the identification results of each key component of the target vehicle.

Step S204, determining the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle according to the recognition result.

Specifically, based on the recognition results, the relative positions of the key components of the target vehicle are determined, and the display area of the target vehicle on the target vehicle picture is determined according to the relative positions of the key components of the target vehicle, and the target vehicle in the display area is extracted at the same time. Determine the vehicle model of the target vehicle.

Further, by analyzing the identification results, the relative positions of the key components of the target vehicle can be determined, and then the display area of the target vehicle on the picture of the target vehicle can be preliminarily determined according to the relative positions of the key components of the target vehicle. By extracting the target vehicle in the display area, the extracted target vehicle is matched with the existing sample vehicle in the database. When the matching is successful, the vehicle model of the target vehicle can be determined.

Step S206, according to the recognition result and the vehicle model, determine the shooting angle of the picture of the target vehicle to be detected.

Specifically, according to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the feature vector matrix corresponding to each key component of the target vehicle is determined. Then, according to the vehicle model of the target vehicle, the baseline feature vector of each key component of the corresponding vehicle is obtained, and the shooting angle of the image of the target vehicle to be detected is determined by comparing the baseline feature vector with the feature vector matrix, and rotating the baseline feature vector.

Among them, by determining the positions of key points of each key component of the target vehicle, extracting the key points on each key point, and then based on the preset sequence of key points, the key point angle vector between any two key points can be calculated and obtained . According to the vehicle model of the target vehicle, a sample vehicle of the same vehicle model can be obtained, and the feature vector of each key component of the sample vehicle can be extracted and determined as the baseline feature vector.

Step S208, according to the vehicle model, obtain the standard relative positional relationship and standard contour line of each key component corresponding to the vehicle.

Specifically, according to the determined vehicle model of the target vehicle, a vehicle of the same vehicle model is determined from the sample vehicles, and the standard relative positional relationship and standard contour line of each key component of the corresponding vehicle are obtained.

Among them, the key components of the sample vehicle include left front light, right front light, left tail light, right tail light, left front door, right front door, left rear door, right rear door, left front wheel, right front wheel, left rear wheel, right rear wheel, front stop Wind glass, rear windshield, hood, trunk lid, left front window, right front window, left rear window, right rear window, left rearview mirror, right rearview mirror, front fender and rear fender, a total of 24 The key components.

Step S210, correcting the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture.

Specifically, the display area of the target vehicle on the target vehicle picture is corrected according to the key components of the sample vehicle, the marked relative positional relationship between the key components, and the determined shooting angle of the target vehicle picture.

Further, correcting the display area of the target vehicle on the target vehicle picture is essentially to realize the deformity correction of the target vehicle. According to the shooting angle of the target vehicle picture, the currently pending target vehicle picture is converted into a target vehicle suitable for analysis. Orientation of the picture to get as much coverage as possible of the display area of the various parts of the vehicle.

Step S212, extracting the contour line of the target vehicle in the corrected display area.

Specifically, the contour line of the target vehicle is obtained by using an edge recognition algorithm to extract the contour line of the target vehicle in the corrected display area.

Further, edge detection is performed on the target vehicle on the target vehicle picture, which specifically includes: 1) filtering: designing a filter to reduce noise; 2) enhancing: using an enhancement algorithm to highlight points with significant changes in grayscale in the field, which can be passed through Calculate the gradient amplitude to complete; 3) Detection: use the shaving amplitude threshold to determine, and detect the edge point; 4) Positioning: accurately determine the position of the edge.

In this embodiment, an edge extraction operator such as the Robert operator or the Sober operator may be used to detect and extract the edge of the target vehicle, thereby obtaining the contour line of the target vehicle.

In step S214, the contour line of the target vehicle is compared with the standard contour line to approximate the edge, and the actual position of each key component of the target vehicle is determined.

Specifically, the edge approximate comparison result is obtained by comparing the contour line of the target vehicle with the standard contour line. Based on the approximate edge comparison results, the relationship between the key components of the target vehicle and the key components of the same vehicle model is established, and then the actual position of each key component of the target vehicle is determined according to the relationship and the standard relative position relationship.

Among them, through the edge approximation comparison and contour recognition, the relationship between the key components of the target vehicle and the key components of the vehicle of the same vehicle model can be established. The position of the key components of the target vehicle is accurately identified, and the actual positions of the key components of the target vehicle that have an associated relationship are obtained.

In the above vehicle component identification method, by identifying the acquired picture of the target vehicle to be detected, the identification results of each key component of the target vehicle picture are obtained, and the display area of the target vehicle on the target vehicle picture and the target vehicle's display area are determined according to the identification results. vehicle model. According to the recognition result and the vehicle model, the shooting angle of the image of the target vehicle to be detected is determined, and according to the vehicle model, the standard relative position relationship and standard contour line of each key component of the corresponding vehicle are obtained. Then, according to the standard relative position relationship and the shooting angle of the target vehicle picture, the display area is corrected to obtain a display area covering all key components of the vehicle as much as possible. By extracting the contour line of the target vehicle in the corrected display area, and performing an approximate edge comparison between the contour line of the target vehicle and the standard contour line, the relationship between the various components can be established. Based on the contour line and the relationship between the components, the actual position of each key component of the target vehicle is further determined, and the accuracy of the identification of the key components of the vehicle is improved. construction.

In one embodiment, as shown in FIG. 3 , the step of recognizing the target image to be detected is the step of acquiring the image of the target vehicle to be detected, recognizing the image of the target vehicle, and obtaining the recognition results of each key component of the image of the target vehicle , which specifically includes the following steps from S302 to S308:

Step S302, acquiring a picture of the target vehicle to be detected.

Before acquiring the image of the target vehicle to be detected, the image to be detected obtained from the terminal device needs to be confirmed to determine whether there is a target vehicle on the image to be detected. When it is determined that there is a target vehicle on the picture to be detected through target detection, the picture to be detected is determined as the picture of the target vehicle to be detected, and the determined picture of the target vehicle to be detected is acquired.

Step S304, acquiring the convolutional neural network model trained on the vehicle pictures of the sample set.

Among them, generating a trained convolutional neural network model specifically includes: obtaining a plurality of standard vehicle pictures that are determined to include vehicles from a database, obtaining a sample set according to the plurality of standard vehicle pictures, and performing the key components of the vehicle pictures in the sample set. callout. Then, the convolutional neural network model is trained according to the labeled sample set vehicle pictures, and the trained convolutional neural network model is generated.

Further, for the steps of labeling the key components of the vehicle pictures in the training sample set, please refer to Figures 4 and 5. Figure 4 provides a schematic diagram of the labeling of the first part of the key components of a standard vehicle, and Figure 5 provides a standard The second part of the vehicle's key components are marked with a schematic diagram. Among them, as shown in Figures 4 and 5, the key components of a standard vehicle include left headlight, right headlight, left taillight, right taillight, left front door, right front door, left rear door, right rear door, left front wheel, right front wheel, left Rear Wheel, Right Rear Wheel, Front Windshield, Rear Windshield, Hood, Tailgate, Left Front Window, Right Front Window, Left Rear Window, Right Rear Window, Left Mirror, Right Mirror, Front Leaf boards and rear fenders.

Step S306, the target vehicle picture is input into the trained convolutional neural network model to identify the target vehicle picture.

Specifically, by inputting the obtained image of the target vehicle into the trained convolutional neural network model, the trained convolutional neural network model is used to identify the image of the target vehicle, and the corresponding key components of the image of the target vehicle are generated. Identify the results.

In step S308, the identification result of each key component of the target vehicle picture is obtained.

Specifically, by obtaining the output results of the trained convolutional neural network model for the target vehicle picture, the identification results of each key component of the target vehicle picture are obtained.

Step 309 further includes: uploading the identification result to the blockchain. The recognition result is uploaded to the blockchain, the user equipment can obtain the image recognition result from the blockchain, and the merchant device can also obtain the image recognition result.

In the above steps, the image of the target vehicle is identified by acquiring the image of the target vehicle to be detected and the convolutional neural network model trained by the image of the vehicle in the sample set, and inputting the image of the target vehicle into the trained convolutional neural network model, The identification results of each key component of the target vehicle picture can be obtained, the user does not need to manually screen the target vehicle picture, the error rate of screening the target vehicle picture is reduced, and the recognition accuracy of each key component of the target vehicle picture is further improved.

In one embodiment, as shown in FIG. 6 , the step of determining the display area on the target vehicle picture and the vehicle model of the target vehicle is to determine the display area of the target vehicle on the target vehicle picture and the target vehicle according to the recognition result. The steps of the vehicle model specifically include the following S602 to S606:

Step S602, based on the identification result, determine the relative positions of the key components of the target vehicle.

Specifically, the identification result of each key component of the target vehicle picture includes the relative positional relationship of each key component of the target vehicle. By analyzing the identification result of each key component of the target vehicle picture, the relative position of each key component of the target vehicle can be determined.

Step S604: Determine the display area of the target vehicle on the image of the target vehicle according to the relative positions of the key components of the target vehicle.

Specifically, according to the relative positions of the key components of the target vehicle, the display position of the target vehicle on the target vehicle picture is preliminarily determined, thereby obtaining the display area of the target vehicle on the target vehicle picture.

Step S606, extract the target vehicle in the display area, and determine the vehicle model of the target vehicle.

Specifically, by extracting the target vehicle in the display area, and matching the target vehicle with different existing sample vehicles in the database, when the matching is successful, the vehicle model of the sample vehicle is obtained, and the vehicle model of the sample vehicle is determined as The vehicle model of the target vehicle.

In the above steps, by using the identification results of the key components of the target vehicle picture, the relative positions of the key components of the target vehicle are determined, and then the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle are determined, without the need for the user to manually treat the detection. The key components of the target vehicle image are marked, reducing manual operations, reducing workload and errors in the process of marking components, and improving the accuracy of marking key components of the vehicle.

In one embodiment, as shown in FIG. 7 , the step of determining the shooting angle of the picture of the target vehicle to be detected, that is, the step of determining the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model, specifically includes the following S702 to S706:

Step S702, according to the identification result of each key component of the target vehicle, obtain the key point angle vector of each key component of the target vehicle, and determine the eigenvector matrix corresponding to each key component of the target vehicle.

Specifically, referring to FIG. 8 , FIG. 8 provides a schematic diagram of the positions of key points of key components of the target vehicle. As shown in FIG. 8 , by analyzing the identification results of each key component of the target vehicle, the position of the key point of each key component is determined, and the key point at the position of each key point is extracted. Based on the preset key point arrangement sequence, the key point angle vector between any two key points is calculated, and the relevant angle vector corresponding to the key point is determined according to the key point angle vector. Then, according to the angle vector of the key point and the corresponding relevant angle vector, the eigenvector matrix corresponding to each key component of the target vehicle is obtained.

Among them, by marking the coordinates of the left headlight of the target vehicle shown in FIG. 8 as (0, 0), and marking the position of the left headlight as the origin of the coordinate system, the coordinate system is established, and the other components included in the coordinate system are sequentially acquired and recorded. The coordinate position of the key point. The preset sequence of key points may be clockwise. Referring to FIG. 9, FIG. 9 provides a schematic diagram of the angle vector of key points of key components of the target vehicle. Referring to FIG. 9, based on the preset sequence of key points, The key point angle vector between the two key points is obtained arbitrarily by calculation, and the key point angle vector of each key component of the target vehicle can be obtained.

Further, a complex index can be used to express the directional relationship between any key point and the current key point. By recording the angle _vector between any key point and the current key point as an , according to the angle vector of the key point, the correlation of the corresponding key point can be determined. The angle vector is:

According to the key point angle vector and the corresponding relevant angle vector, the eigenvector matrix corresponding to each key component of the target vehicle can be obtained, as follows:

Step S704, according to the vehicle model of the target vehicle, obtain the baseline feature vector of each key component of the corresponding vehicle.

Specifically, referring to FIG. 10 , FIG. 10 provides a schematic diagram of a baseline feature vector of a sample vehicle. According to the vehicle model of the target vehicle, a sample vehicle of the same vehicle model can be obtained, and the baseline feature vector of each key component of the sample vehicle can be extracted.

Further, by vectorizing the 3D model of the model corresponding to each sample vehicle, take the left headlight, right headlight, left taillight, right taillight, left front door, right front door, left rear door, right rear door, left front wheel, right Front Wheel, Left Rear Wheel, Right Rear Wheel, Front Windshield, Rear Windshield, Hood, Tailgate, Left Front Window, Right Front Window, Left Rear Window, Right Rear Window, Left Mirror, Right Rear View Mirrors, front fenders are organic rear fenders, and the spatial positions of each component take their respective center points. With the left headlight as the spatial coordinate (0, 0, 0), the positions of different components are recorded respectively.

Wherein, any part is recorded as: P _obj =(x, y, z).

Then left front light, right front light, left tail light, right tail light, left front door, right front door, left rear door, right rear door, left front wheel, right front wheel, left rear wheel, right rear wheel, front windshield, rear windshield , hood, trunk lid, left front window, right front window, left rear window, right rear window, left rearview mirror, right rearview mirror, front fender, and rear fender. The vehicle part is a 24-dimensional feature vector ,As follows:

In one embodiment, since the vehicle is actually 3D, but the vehicle photo obtained by taking the vehicle is flat, the calculation can be performed based on the plane projection of the feature vector of the key components of the vehicle at any angle, and the vector of each key component included in the vehicle can be obtained. Projection on the XY plane. Different shooting angles have different projections, which can provide a basis for the subsequent judgment of the shooting angle through the position of the projection.

Step S706, compare the baseline feature vector with the feature vector matrix, rotate the baseline feature vector, and determine the shooting angle of the image of the target vehicle to be detected.

Specifically, the comparison result is obtained by extracting the horizontal eigenvector and the vertical eigenvector of the eigenvector matrix, and comparing the horizontal eigenvector and the vertical eigenvector with the baseline eigenvector. Then, based on the comparison result, the rotation angle of the baseline feature vector is determined. Rotate the baseline feature vector according to the rotation angle, and calculate the geometric mean of the angle vector of the common key points during the rotation of the baseline vector, where the common key point is the common key point of the feature vector matrix and the baseline feature vector. When the geometric mean value of the angle vector of the common key point reaches the preset threshold, the shooting angle of the image of the target vehicle to be detected is obtained.

Further, as shown in FIG. 11 , FIG. 11 provides a schematic diagram of a vector comparison of a vehicle component identification method. Due to the different shooting angles, the projected position of the key points of the target vehicle on the target vehicle image is inconsistent with the projected position of the key points of the sample vehicle. By comparing the baseline feature vector of the sample vehicle with the horizontal feature vector of the key points of the target vehicle and The vertical feature vectors are compared to obtain the vector comparison diagram of the vehicle component identification method as shown in FIG. 11 .

Among them, since the baseline feature vector does not coincide with the horizontal feature vector and the vertical feature vector, the baseline feature vector can be rotated according to the rotation angle determined by the vector comparison. By rotating the baseline feature vector, and calculating the geometric mean of the angle vectors of the common keypoints during the rotation of the baseline vector. When the baseline eigenvector is rotated so that the baseline vector is gradually approximated to the horizontal eigenvector and the vertical eigenvector, the geometric mean of the angle vectors of the common key points gradually decreases. When the geometric mean of the angle vectors of the common key points takes the minimum value, the shooting angle of the target vehicle picture is obtained.

Among them, the calculation formula of the geometric mean of the angle vectors of the common key points is as follows:

表示目标车辆的公共关键点的角度向量，

表示样本车辆的公共关键点的角度向量，基于上述计算公式，可得到公共关键点的角度向量的几何均值。在旋转过程中，当公共关键点的角度向量的几何均值取最小值时，得到目标车辆图片的拍摄角度。in,

the angle vector representing the common keypoints of the target vehicle,

The angle vector representing the common key point of the sample vehicle, based on the above calculation formula, the geometric mean of the angle vector of the common key point can be obtained. During the rotation process, when the geometric mean of the angle vectors of the common key points takes the minimum value, the shooting angle of the target vehicle picture is obtained.

In the above steps, according to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the feature vector matrix corresponding to each key component of the target vehicle is determined. According to the vehicle model of the target vehicle, obtain the baseline feature vector of each key component of the corresponding vehicle, compare the baseline feature vector with the feature vector matrix, rotate the baseline feature vector, and quickly determine the shooting angle of the target vehicle picture, so as to judge The quality of the image shooting is prepared for the subsequent vehicle image quality inspection, and the work efficiency of vehicle image inspection is improved.

In one embodiment, after obtaining the approximate edge comparison result, the method further includes: determining the key deformation components of the target vehicle based on the approximate edge comparison results, and calculating the deformation rate of the key deformation components.

Specifically, based on the edge approximation comparison result, determine the deformation key components of the target vehicle, and determine the damaged key components. Then, by calculating the deformation rate of the deformation key components, and comparing the deformation key components with the corresponding key components of the sample vehicle, the damage status of the key components is quantified based on the deformation rate.

In the above steps, based on the edge approximation comparison results, determine the deformation key components of the target vehicle, and calculate the deformation rate of the deformation key components. The damage status is quantified to better determine the damage status of the target vehicle, timely and accurately realize vehicle damage assessment, and improve work efficiency.

It should be understood that although the steps in the flowcharts of FIGS. 2 , 3 , 6 and 7 are sequentially displayed according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order, and these steps may be performed in other orders. Moreover, at least a part of the steps in FIG. 2 , FIG. 3 , FIG. 6 and FIG. 7 may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed and completed at the same moment, but may be performed at different moments. The execution order of these sub-steps or phases is not necessarily performed sequentially, but may be performed alternately or alternately with other steps or at least a part of sub-steps or phases of other steps.

In one embodiment, as shown in FIG. 12, a vehicle component identification device is provided, including: a recognition result generation module 1202, a display area determination module 1204, a shooting angle determination module 1206, a first acquisition module 1208, and a display area correction Module 1210, target vehicle contour extraction module 1212, and key component actual position determination module 1214, wherein:

The identification result generation module 1202 is used to obtain the picture of the target vehicle to be detected, identify the picture of the target vehicle, obtain the identification result of each key component of the target vehicle picture, upload the identification result to the blockchain, and the user equipment can download the identification result from the region. The image recognition result is obtained in the blockchain, and the merchant device can also obtain the image recognition result.

The display area determination module 1204 is configured to determine the display area of the target vehicle on the image of the target vehicle and the vehicle model of the target vehicle according to the recognition result.

The shooting angle determination module 1206 is configured to determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model.

The first obtaining module 1208 is configured to obtain the standard relative positional relationship and standard outline of each key component of the corresponding vehicle according to the vehicle model.

The display area correction module 1210 is used to correct the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture.

The target vehicle contour line extraction module 1212 is used to extract the contour line of the target vehicle in the corrected display area.

The actual position determination module 1214 of key components is configured to perform an edge approximation comparison between the contour line of the target vehicle and the standard contour line to determine the actual position of each key component of the target vehicle.

In the above vehicle component identification device, by identifying the acquired image of the target vehicle to be detected, the identification result of each key component of the target vehicle image is obtained, and the display area of the target vehicle on the target vehicle image and the target vehicle's display area are determined according to the identification result. vehicle model. According to the recognition result and the vehicle model, the shooting angle of the image of the target vehicle to be detected is determined, and according to the vehicle model, the standard relative position relationship and standard contour line of each key component of the corresponding vehicle are obtained. Then, according to the standard relative position relationship and the shooting angle of the target vehicle picture, the display area is corrected to obtain a display area covering all key components of the vehicle as much as possible. By extracting the contour line of the target vehicle in the corrected display area, and performing an approximate edge comparison between the contour line of the target vehicle and the standard contour line, the relationship between the various components can be established. Based on the contour line and the relationship between the components, the actual position of each key component of the target vehicle is further determined, and the accuracy of the identification of the key components of the vehicle is improved.

In one embodiment, the shooting angle determination module is further used for:

According to the identification results of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the eigenvector matrix corresponding to each key component of the target vehicle is determined; Baseline feature vector of each key component; compare the baseline feature vector with the feature vector matrix, rotate the baseline feature vector, and determine the shooting angle of the image of the target vehicle to be detected.

In the above shooting angle determination module, according to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the feature vector matrix corresponding to each key component of the target vehicle is determined. According to the vehicle model of the target vehicle, obtain the baseline feature vector of each key component of the corresponding vehicle, compare the baseline feature vector with the feature vector matrix, rotate the baseline feature vector, and quickly determine the shooting angle of the target vehicle picture, so as to judge The quality of the image shooting is prepared for the subsequent vehicle image quality inspection, and the work efficiency of vehicle image inspection is improved.

In one embodiment, the recognition result generation module is further used for:

Obtain the image of the target vehicle to be detected; obtain the convolutional neural network model trained by the vehicle image of the sample set; input the image of the target vehicle into the trained convolutional neural network model to identify the image of the target vehicle; obtain the key points of the image of the target vehicle The identification result of the part.

In the above recognition result generation module, the image of the target vehicle is obtained by acquiring the image of the target vehicle to be detected and the convolutional neural network model trained by the vehicle image in the sample set, and inputting the image of the target vehicle into the trained convolutional neural network model. By performing identification, the identification results of key components in the target vehicle picture can be obtained, without the need for the user to manually screen the target vehicle picture, reducing the error rate of screening the target vehicle picture, and further improving the identification accuracy of each key component in the target vehicle picture.

In one embodiment, the display area determination module is further configured to:

Based on the recognition results, determine the relative positions of the key components of the target vehicle; determine the display area of the target vehicle on the target vehicle picture according to the relative positions of the key components of the target vehicle; extract the target vehicle in the display area, and determine the vehicle of the target vehicle model.

The above-mentioned display area determination module determines the relative position of each key component of the target vehicle by using the identification result of each key component of the target vehicle picture, and then determines the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle, without the need for a user Manually label the key components of the target vehicle image to be detected, reduce manual operations, reduce workload and errors in the process of labeling components, and improve the accuracy of labeling key components of the vehicle.

In one embodiment, a vehicle component identification device is provided, further comprising a deformation rate calculation module for:

Based on the edge approximation comparison results, the deformation key components of the target vehicle are determined, and the deformation rate of the deformation key components is calculated.

The above vehicle component identification device, based on the edge approximate comparison result, determines the deformation key components of the target vehicle, calculates the deformation rate of the deformation key components, and determines whether the damaged key components are damaged on the basis of the deformation rate. Calculate and quantify the damage condition, better judge the damage condition of the target vehicle, timely and accurately realize the vehicle damage assessment, and improve work efficiency.

For the specific definition of the vehicle component identification device, reference may be made to the above definition of the vehicle component identification method, which will not be repeated here. Each module in the above-mentioned vehicle component identification device may be implemented in whole or in part by software, hardware, or a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided, and the computer device may be a server, and its internal structure diagram may be as shown in FIG. 13 . The computer device includes a processor, memory, a network interface, and a database connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used to store the relative positions of the key components of the target vehicle and the contour lines of the target vehicle, the marked relative position relationship of the key components of the sample vehicles of the same vehicle model, and the standard contour lines. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by a processor, implements a vehicle component identification method.

Those skilled in the art can understand that the structure shown in FIG. 13 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps in each of the foregoing method embodiments are implemented.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, implements the steps in each of the foregoing method embodiments.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other medium used in the various embodiments provided in this application may include non-volatile and/or volatile memory. Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features It is considered to be the range described in this specification.

The above-mentioned embodiments only represent several embodiments of the present application, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the scope of protection of the patent of the present application shall be subject to the appended claims.

Claims (10)

1. A vehicle component identification method, the method comprising: Obtaining a picture of the target vehicle to be detected, identifying the picture of the target vehicle, and obtaining the identification result of each key component of the picture of the target vehicle; Determine the display area of the target vehicle on the target vehicle picture and the vehicle model of the target vehicle according to the recognition result; Determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model; According to the vehicle model, obtain the standard relative positional relationship and standard contour line of each key component of the corresponding vehicle; Correcting the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture; extracting the contour line of the target vehicle in the display area after correction; The contour line of the target vehicle is compared with the standard contour line by edge approximation to determine the actual position of each key component of the target vehicle. 2. The method according to claim 1, wherein obtaining a picture of the target vehicle to be detected, identifying the picture of the target vehicle, and obtaining a recognition result of each key component of the picture of the target vehicle, comprising: Obtain a picture of the target vehicle to be detected; Obtain the convolutional neural network model trained by the sample set of vehicle images; Input the image of the target vehicle into the trained convolutional neural network model, and identify the image of the target vehicle; Obtain the identification results of each key component of the target vehicle picture; Also includes: Upload the identification result to the blockchain. 3 . The method according to claim 2 , wherein determining the display area of the target vehicle on the picture of the target vehicle and the vehicle model of the target vehicle according to the recognition result, comprising: 3 . Based on the identification result, determine the relative positions of the key components of the target vehicle; Determine the display area of the target vehicle on the picture of the target vehicle according to the relative positions of the key components of the target vehicle; The target vehicle in the display area is extracted, and the vehicle model of the target vehicle is determined. 4 . The method according to claim 2 , wherein determining the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model, comprising: 5 . According to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the eigenvector matrix corresponding to each key component of the target vehicle is determined; According to the vehicle model of the target vehicle, obtain the baseline feature vector of each key component of the corresponding vehicle; The baseline feature vector is compared with the feature vector matrix, the baseline feature vector is rotated, and the shooting angle of the image of the target vehicle to be detected is determined. 5. The method according to any one of claims 1 to 4, wherein the approximate edge comparison between the contour line of the target vehicle and the standard contour line is performed to determine the key points of the target vehicle. The actual location of the part, including: Performing an approximate edge comparison between the contour line of the target vehicle and the standard contour line to obtain an approximate edge comparison result; establishing an association relationship between the key components of the target vehicle and the key components of vehicles of the same vehicle model based on the edge approximation comparison result; According to the association relationship and the standard relative position relationship, the actual position of each key component of the target vehicle is determined. 6 . The method according to claim 4 , wherein according to the identification result of each key component of the target vehicle, the key point angle vector of each key component of the target vehicle is obtained, and the key point angle vector of each key component of the target vehicle is determined, and the key point angle vector of each key component of the target vehicle is determined. 7 . The eigenvector matrix corresponding to each key component of , including: According to the identification result of each key component of the target vehicle, determine the key point position of each key component; Extracting key points at the positions of each of the key points, and calculating a key point angle vector between any two key points based on a preset arrangement order; According to the key point angle vector, determine the relevant angle vector corresponding to the key point; According to the key point angle vector and the corresponding relevant angle vector, a feature vector matrix corresponding to each key component of the target vehicle is obtained. 7. The method according to claim 4, characterized in that, comparing the baseline feature vector with the feature vector matrix, rotating the baseline feature vector, and determining the Shooting angles, including: extracting horizontal eigenvectors and vertical eigenvectors of the eigenvector matrix; Comparing the horizontal feature vector and the vertical feature vector with the baseline feature vector to obtain a comparison result; Based on the comparison result, determine the rotation angle of the baseline feature vector; Rotate the baseline feature vector according to the rotation angle, and calculate the geometric mean of the angle vector of the common key point in the process of rotating the baseline vector; the common key point is the common feature vector matrix and the baseline feature vector. key point; When the geometric mean of the angle vectors of the common key points reaches a preset threshold, the shooting angle of the picture of the target vehicle to be detected is obtained. 8. A vehicle component identification device, characterized in that the device comprises: A recognition result generation module, configured to obtain a picture of the target vehicle to be detected, identify the picture of the target vehicle, and obtain the recognition results of each key component of the picture of the target vehicle; a display area determination module, configured to determine the display area of the target vehicle on the picture of the target vehicle and the vehicle model of the target vehicle according to the recognition result; a shooting angle determination module, configured to determine the shooting angle of the picture of the target vehicle to be detected according to the recognition result and the vehicle model; a first acquisition module, configured to acquire, according to the vehicle model, standard relative positional relationships and standard contour lines of each key component of the corresponding vehicle; a display area correction module, configured to correct the display area according to the standard relative position relationship and the shooting angle of the target vehicle picture; a target vehicle contour line extraction module, used for extracting the corrected contour line of the target vehicle in the display area; The actual position determination module of key components is used to compare the contour line of the target vehicle with the standard contour line by approximating the edge to determine the actual position of each key component of the target vehicle. 9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when the processor executes the computer program . 10. A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images