Method and device for constructing vehicle three-dimensional model, electronic equipment and medium

Abstract

The present disclosure provides a vehicle three-dimensional model construction method and device, electronic equipment and medium, relates to the technical field of image processing, and particularly relates to the technical field of intelligent traffic, intelligent parking and cloud service. The specific implementation scheme is: obtaining a candidate collection image associated with target license plate information, and determining vehicle attribute information of a candidate vehicle in the candidate collection image; selecting a target vehicle matching the target license plate information from the candidate vehicle according to the vehicle attribute information, and taking a candidate collection image to which the target vehicle belongs as a target collection image; and constructing a three-dimensional model of the target vehicle according to the target collection image. The present disclosure achieves the effect of improving the accuracy of vehicle three-dimensional model construction.

Description

Technical Field

[0001] This disclosure relates to the field of image processing technology, and more particularly to the fields of intelligent transportation, smart parking and cloud services technology, and especially to a method, apparatus, electronic device and medium for constructing a three-dimensional vehicle model. Background Technology

[0002] With the development of technology and the improvement of people's living standards, more and more people are choosing to drive private cars. Correspondingly, more and more cameras are being installed on the roads to replace human traffic supervision, such as parking space supervision or traffic violation supervision.

[0003] To improve the accuracy of monitoring, traffic surveillance using cameras is now based on 3D models of vehicles. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and medium for improving the accuracy of building 3D models of vehicles.

[0005] According to one aspect of this disclosure, a method for constructing a three-dimensional model of a vehicle is provided, comprising:

[0006] Acquire candidate images associated with the target license plate information, and determine the vehicle attribute information of the candidate vehicles in the candidate images;

[0007] Based on the vehicle attribute information, a target vehicle matching the target license plate information is selected from the candidate vehicles, and the candidate image of the target vehicle is used as the target image.

[0008] A three-dimensional model of the target vehicle is constructed based on the target image.

[0009] According to another aspect of this disclosure, an apparatus for constructing a three-dimensional model of a vehicle is provided, comprising:

[0010] The vehicle attribute information determination module is used to acquire candidate images associated with the target license plate information and determine the vehicle attribute information of the candidate vehicles in the candidate images.

[0011] The target image acquisition module is used to select a target vehicle that matches the target license plate information from the candidate vehicles based on the vehicle attribute information, and to use the candidate acquisition image of the target vehicle as the target acquisition image.

[0012] The model building module is used to build a three-dimensional model of the target vehicle based on the target acquired image.

[0013] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0014] At least one processor; and

[0015] A memory that is communicatively connected to at least one processor; wherein,

[0016] The memory stores instructions that can be executed by at least one processor to enable the at least one processor to perform any of the methods of this disclosure.

[0017] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform any of the methods of this disclosure.

[0018] According to another aspect of this disclosure, a computer program product is provided, including a computer program and a method for the computer program to be executed by a processor according to any of the methods disclosed herein.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0020] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0021] Figure 1 This is a flowchart of a method for constructing a three-dimensional vehicle model according to an embodiment of this disclosure;

[0022] Figure 2 This is a flowchart of another method for constructing a three-dimensional vehicle model according to embodiments of this disclosure;

[0023] Figure 3 This is a flowchart of another method for constructing a three-dimensional vehicle model according to embodiments of this disclosure;

[0024] Figure 4 This is a flowchart of some three-dimensional model verification methods disclosed in the embodiments of this disclosure;

[0025] Figure 5 This is a schematic diagram of the structure of a vehicle three-dimensional model construction device disclosed in the embodiments of this disclosure;

[0026] Figure 6 This is a block diagram of an electronic device used to implement the method for constructing a three-dimensional vehicle model disclosed in the embodiments of this disclosure. Detailed Implementation

[0027] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0028] Different types of cameras on the road are installed at varying heights and angles, capturing a large number of two-dimensional images of the monitored scene during routine surveillance. These images can be analyzed to monitor traffic, such as parking violations, speeding, and running red lights. However, due to the numerous blind spots and visual errors in two-dimensional images, traffic surveillance based on them suffers from relatively low accuracy.

[0029] To improve the accuracy of surveillance, current traffic monitoring using cameras is based on 3D vehicle models. The construction of these 3D vehicle models typically involves: capturing images of the vehicle individually and then building a 3D model based on those images.

[0030] However, the above method requires separate image acquisition of the vehicle, resulting in high image acquisition costs and low model building efficiency. Furthermore, the number of images and shooting angles obtained through this method are limited, leading to similarly low accuracy in 3D model construction.

[0031] Figure 1 This is a flowchart illustrating methods for constructing 3D vehicle models according to embodiments of this disclosure. These embodiments are applicable to the situation of performing 3D modeling of vehicles from acquired images. The methods in this embodiment can be executed by a vehicle 3D model construction apparatus disclosed in these embodiments. The apparatus can be implemented using software and / or hardware and can be integrated into any electronic device with computing capabilities.

[0032] like Figure 1 As shown, the method for constructing a three-dimensional vehicle model disclosed in this embodiment may include:

[0033] S101. Obtain candidate images associated with the target license plate information, and determine the vehicle attribute information of the candidate vehicles in the candidate images.

[0034] The target license plate information refers to the license plate information that matches the vehicle for which a 3D model needs to be constructed. Candidate images represent vehicle images captured by image acquisition equipment based on the license plate information; optionally, candidate images can be captured by cameras installed on various roads. Vehicle attribute information represents the appearance features of the candidate vehicles, including but not limited to vehicle size, vehicle type, and vehicle color.

[0035] Cameras installed on roads capture images of vehicles in the designated area and use image processing algorithms to identify license plate information within these images, such as OCR (Optical Character Recognition) algorithms. These cameras can be installed on multiple roads, with different angles and heights on each road.

[0036] The captured image is used as a candidate captured image, and the candidate captured image is associated with the recognized license plate information and stored. The association relationship between the license plate information and the candidate captured image is established. For example, the license plate information and the candidate captured image are associated and stored in the storage medium built into the camera device, or the license plate information and the candidate captured image are associated and stored in the cloud server. This embodiment does not limit the specific method of associating and storing the license plate information and the candidate captured image.

[0037] Optionally, after the camera device identifies the vehicle's license plate information, it tracks the vehicle using a target tracking algorithm to obtain continuous captured images as candidate captured images. In other words, in this embodiment, it is not necessary for each frame of candidate captured images to contain the vehicle's license plate information.

[0038] In one implementation of S101, when it is necessary to construct a 3D model of any vehicle whose license plate information matches, the license plate information is used as the target license plate information. The correlation between the target license plate information and the license plate information is matched with the candidate acquisition images to determine the candidate acquisition images associated with the target license plate information, and the candidate acquisition images are retrieved from the storage location. For example, assuming the target license plate information is "AABBCC", and there is a correlation between the license plate information "AABBCC" and the candidate acquisition image X, then the candidate acquisition image X is used as the candidate acquisition image associated with the target license plate information "AABBCC".

[0039] Vehicle identification is performed on candidate images to determine at least one candidate vehicle. Then, vehicle attribute detection is performed on the candidate vehicles to determine their attribute information. Vehicle attribute detection includes, but is not limited to, the following methods:

[0040] A. Determine the vehicle type and color of candidate vehicles in the candidate acquired images using a deep learning model. B. Determine the vehicle size of the candidate vehicles using the corresponding 3D acquired images. In this embodiment, the vehicle attribute information may include only vehicle type and color, only vehicle size, or all three. Accordingly, if the vehicle attribute information only includes vehicle type and color, then method A is used to determine the vehicle attribute information; if the vehicle attribute information only includes vehicle size, then method B is used to determine the vehicle attribute information; if the vehicle attribute information includes all three, then methods A and B are used to determine the vehicle attribute information.

[0041] By acquiring candidate images associated with the target license plate information and determining the vehicle attribute information of the candidate vehicles in the candidate images, the effect of obtaining vehicle attribute information was achieved, laying a data foundation for subsequent vehicle filtering based on vehicle attribute information.

[0042] S102. Based on the vehicle attribute information, select the target vehicle that matches the target license plate information from the candidate vehicles, and use the candidate acquisition image of the target vehicle as the target acquisition image.

[0043] In one implementation, standard attribute information of the target vehicle is obtained, and the vehicle attribute information of the candidate vehicles is compared with the standard attribute information. Then, the target vehicle is determined from the candidate vehicles based on the comparison result, and the candidate acquisition image to which the target vehicle belongs is used as the target acquisition image.

[0044] Optionally, the vehicle dimensions of the candidate vehicles are compared with those of the standard vehicles, the vehicle types of the candidate vehicles are compared with those of the standard vehicles, and the vehicle colors of the candidate vehicles are compared with those of the standard vehicles. Candidate vehicles whose attribute information simultaneously meets the requirements of standard vehicle dimensions, standard vehicle types, and standard vehicle colors are then selected as target vehicles.

[0045] In another implementation, the vehicle size, vehicle type, and vehicle color of each candidate vehicle are statistically analyzed. The vehicle size with the most statistically analyzed values ​​is selected as the target vehicle size, the vehicle type with the most statistically analyzed values ​​is selected as the target vehicle type, and the vehicle color with the most statistically analyzed values ​​is selected as the target vehicle color. Then, the candidate vehicle that simultaneously meets the target vehicle size, target vehicle type, and target vehicle color is selected as the target vehicle, and the candidate image to which the target vehicle belongs is selected as the target image.

[0046] By selecting the target vehicle that matches the target license plate information from the candidate vehicles based on vehicle attribute information, and using the candidate image of the target vehicle as the target image, the effect of excluding incorrect candidate vehicles based on vehicle attribute information is achieved.

[0047] S103. Construct a three-dimensional model of the target vehicle based on the target acquisition image.

[0048] In one implementation, the target image is segmented to obtain the mask region and key point positions of the target vehicle in the target image. Then, a three-dimensional model of the target vehicle is constructed based on the target image, the mask region, and the key positions to generate a three-dimensional model of the target vehicle.

[0049] Optionally, the target acquisition image, mask region, and key locations can be used as input to the SFM (Structure From Motion) algorithm to generate a 3D model of the target vehicle.

[0050] After generating a 3D model of the target vehicle, the 3D model is associated with and stored along with the target license plate information. This 3D model can be used for 3D scene reconstruction, such as parking monitoring or traffic violation monitoring. For example, when the target license plate information of a target vehicle is detected, the 3D model of the target vehicle is retrieved from the data object storage space based on the target license plate information, thereby accurately analyzing the positional relationship between the target vehicle and space based on the 3D model of the target vehicle.

[0051] This disclosure achieves the effect of constructing a 3D model of a vehicle using roadside camera equipment by acquiring candidate images associated with the target license plate information and determining the vehicle attribute information of the candidate vehicles in the candidate images. This method acquires a large number of images from various angles, improving the accuracy of the 3D model construction. Furthermore, it eliminates the need for separate image acquisition of the target vehicle, significantly reducing image acquisition costs and improving model construction efficiency. By selecting the target vehicle that matches the target license plate information from the candidate vehicles based on the vehicle attribute information, and using the candidate images of the target vehicle as the target acquisition images, the 3D model of the target vehicle is constructed based on the target acquisition images. Since the target vehicle is determined by filtering candidate vehicles based on vehicle attribute information, problems such as cloned license plates or incorrect captures are avoided, ensuring the accuracy of the target vehicle and further improving the accuracy of the 3D model construction of the target vehicle.

[0052] Figure 2 This is a flowchart of another method for constructing a three-dimensional vehicle model according to the embodiments of this disclosure. It is further optimized and extended based on the above technical solutions and can be combined with the above optional implementation methods.

[0053] like Figure 2 As shown, the method for constructing a three-dimensional vehicle model disclosed in this embodiment may include:

[0054] S201. Obtain candidate images associated with the target license plate information.

[0055] S202. Determine the target acquisition camera based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image.

[0056] The cameras are installed on the road to capture images. Each camera number identifies its corresponding camera; that is, any camera number can be used to determine the camera it belongs to. When storing candidate images, in addition to associating them with license plate information, they are also associated with the camera number of the target camera used to capture the candidate images.

[0057] In one implementation, the camera number associated with the candidate acquired image is obtained, and the camera number is matched with the association between the camera number and the acquired camera to determine the target acquired camera associated with the camera number.

[0058] For example, suppose the camera number associated with candidate image A is "12345", and the acquisition camera associated with camera number "12345" is acquisition camera B, then acquisition camera B will be the target acquisition camera.

[0059] S203. Based on the camera parameters of the target acquisition camera, convert the candidate acquisition images into three-dimensional acquisition images.

[0060] Camera parameters include, but are not limited to, camera extrinsic and camera intrinsic parameters. Camera intrinsic parameters represent the camera's internal parameters, including but not limited to focal length, distortion coefficients, tangential distortion, radial distortion, and camera lens distortion. Camera extrinsic parameters include the camera's position and attitude information acquired in the world coordinate system.

[0061] In one implementation, the three-dimensional coordinates of pixels in the candidate acquired image are determined using the following formula:

[0062] Z×W=K×P×V

[0063] Where W represents the image coordinates of a pixel in the candidate acquired image, K represents the camera intrinsic parameter, P represents the camera extrinsic parameter, V represents the three-dimensional coordinates of a pixel in the candidate acquired image, and Z represents the depth information of a pixel in the candidate acquired image, which can be calculated based on the camera extrinsic parameter P.

[0064] Based on the calculated 3D coordinates of pixels in the candidate acquisition images, a 3D acquisition image corresponding to the candidate acquisition image is generated.

[0065] S204. Based on the 3D acquired images, determine the vehicle attitude angles and vehicle dimensions of the candidate vehicles in the world coordinate system.

[0066] Wherein, vehicle attitude angles represent the attitude angles of the candidate vehicle around the X-axis, Y-axis, and Z-axis in the world coordinate system. Vehicle dimensions include, but are not limited to, at least one of the candidate vehicle's length, width, and height values ​​in the world coordinate system.

[0067] In one implementation, target detection is performed on the 3D acquired image to determine the 3D detection box (3D-Boxer) of the candidate vehicle in the 3D acquired image, wherein the 3D detection box is a rectangular detection box.

[0068] The vehicle attitude angles in the world coordinate system are determined based on the angles between the 3D detection bounding box and the X-axis, Y-axis, and Z-axis. Furthermore, the length, width, and height of the candidate vehicle in the world coordinate system are determined based on the length and scaling of the 3D detection bounding box.

[0069] This disclosure determines the target acquisition camera based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image, and the candidate acquisition image is converted into a three-dimensional acquisition image according to the camera parameters of the target acquisition camera, thereby realizing the effect of converting the two-dimensional candidate acquisition image into a three-dimensional acquisition image, laying the foundation for subsequent determination of vehicle attitude angle and vehicle size; by determining the vehicle attitude angle and vehicle size of the candidate vehicle in the world coordinate system based on the three-dimensional acquisition image, the effect of detecting three-dimensional vehicle attribute information based on the three-dimensional acquisition image is realized, expanding the dimensional diversity of vehicle attribute information.

[0070] S205. Determine the average size based on the vehicle size of the candidate vehicles, and determine at least one size range based on the average size and at least one size threshold.

[0071] The vehicle dimensions include at least one of length, width, and height. The size and number of size thresholds can be set according to actual business needs.

[0072] In one implementation, an average size is calculated based on the vehicle dimensions of each candidate vehicle to determine the average size, wherein the average size includes at least one of the average length, average width, and average height. Starting from the average size and using at least one size threshold as a step size, at least one size interval is formed by expanding in two directions: smaller than the average size and larger than the average size, until the size interval covers the smallest and largest vehicle dimensions.

[0073] The above implementation method will be explained using vehicle dimensions as an example. Assume the lengths of candidate vehicles A, B, C, and D are 4m, 5.5m, 4.5m, and 6m respectively, with an average calculated length of 5m. Assume the size threshold is set to 0.1m. Starting from 5m and incrementing by 0.1m, the size intervals [4.9m, 5m] and [5m, 5.1m] are obtained. This is then expanded by increments of 0.1m to obtain size intervals [4.8m, 4.9m] and [5.1m, 5.2m], until the obtained size intervals cover both 4m and 6m.

[0074] S206. Based on the number of vehicle sizes included in the size range, select the target vehicle that matches the target license plate information from the candidate vehicles.

[0075] In one implementation, the number of vehicle sizes contained in each size range is counted, and the target vehicle matching the target license plate information is selected from the candidate vehicles based on the size range containing the largest number of vehicle sizes.

[0076] This disclosure determines the target vehicle based on the distribution statistics of vehicle size by determining the average size of candidate vehicles, and by determining at least one size range based on the average size and at least one size threshold. Based on the number of vehicle sizes included in the size range, the target vehicle matching the target license plate information is selected from the candidate vehicles. This achieves the effect of determining the target vehicle based on the distribution statistics of vehicle size, avoiding problems such as cloned vehicles and false captures, and ensuring the accuracy of determining the target vehicle.

[0077] Optionally, S206 includes:

[0078] The size range containing the most vehicle sizes is selected as the target size range, and the vehicle sizes contained within the target size range are selected as the target vehicle sizes; the candidate vehicles to which the target vehicle sizes belong are selected as the target vehicles.

[0079] For example, suppose there are four size ranges: size range 1, size range 2, size range 3, and size range 4. Size range 1 contains 20 vehicle sizes, size range 2 contains 25 vehicle sizes, size range 3 contains 15 vehicle sizes, and size range 4 contains 18 vehicle sizes. Taking size range 2 as the target size range, assuming that the vehicle sizes in size range 2 include length values ​​of 5m, 5.1m, and 4.9m, then length values ​​of 5m, 5.1m, and 4.9m are selected as target vehicle sizes, and candidate vehicles with length values ​​of 5m, 5.1m, and 4.9m are selected as target vehicles.

[0080] By using the size range containing the most vehicle sizes as the target size range, and the vehicle sizes contained within the target size range as the target vehicle sizes, the candidate vehicles to which the target vehicle sizes belong are then identified as the target vehicles. Since the target size range contains the most vehicle sizes, the probability value of the target vehicle size contained within the target size range belonging to the target vehicle is the highest. This approach minimizes issues such as cloned vehicles and false captures, ensuring the accuracy of identifying the target vehicle.

[0081] S207. Take the candidate acquisition image of the target vehicle as the target acquisition image, and construct a three-dimensional model of the target vehicle based on the target acquisition image.

[0082] In one implementation, the target image is segmented to determine the mask region and key point positions of the target vehicle in the target image, and then a three-dimensional model of the target vehicle is constructed based on the target image, the mask region, and the key point positions.

[0083] Optionally, S207 includes the following steps A, B, and C:

[0084] A. Determine the first filter image from the target acquisition image based on the vehicle attitude angle of the target vehicle in the target acquisition image and the first attitude angle spacing.

[0085] In one embodiment, the target acquisition image is filtered to determine the first filtered image based on the first attitude angle spacing and the vehicle attitude angle of the target vehicle in the target acquisition image.

[0086] For example, assuming the first attitude angle interval is 60 degrees, the vehicle attitude angle of any target vehicle is selected as the starting point, and then the vehicle attitude angle is filtered at intervals of 60 degrees in the X-axis, Y-axis and Z-axis directions respectively to determine the first filtered attitude angle, and the target acquisition image to which the first filtered attitude angle belongs is used as the first filtered image.

[0087] B. Perform target segmentation on the first filtered image to determine the first mask region and the first key point location of the target vehicle in the first filtered image.

[0088] The first mask region refers to the occlusion area covering the target vehicle in the first filtered image, which can be used to extract the region of interest of the target vehicle. The first key point location refers to the location of the target vehicle's feature points in the first filtered image, including but not limited to the tire position, rearview mirror position, door handle position, and sunroof position, etc.

[0089] In one implementation, a target segmentation algorithm is used to segment the first selected image to determine the first mask region and the first key point position of the target vehicle in the first selected image.

[0090] C. Based on the first selected image, the first mask area, and the location of the first key point, construct a three-dimensional model of the target vehicle in the first selected image to generate a three-dimensional model.

[0091] In one implementation, the first screened image, the first mask region, and the first key point location are used as inputs to the SFM algorithm, and the algorithm is executed to generate a three-dimensional model of the target vehicle.

[0092] By determining the first selected image from the target acquired image based on the vehicle attitude angle and the first attitude angle interval of the target vehicle in the target acquired image, the first selected image is segmented to determine the first mask region and the first key point position of the target vehicle in the first selected image. Based on the first selected image, the first mask region, and the first key point position, a 3D model of the target vehicle in the first selected image is constructed to generate a 3D model. Since the first selected image is selected from the target acquired image based on the first attitude angle interval, and then 3D modeling is performed based on the first selected image, the problem of excessive data volume affecting the efficiency of 3D modeling due to 3D modeling based on all target acquired images is avoided. Furthermore, the target vehicles in the first selected image selected based on the first attitude angle interval have different vehicle attitude angles, thus ensuring the accuracy of the 3D modeling results.

[0093] Figure 3 This is a flowchart of another method for constructing a three-dimensional vehicle model according to the embodiments of this disclosure. It is further optimized and extended based on the above technical solutions and can be combined with the above optional implementation methods.

[0094] like Figure 3 As shown, the method for constructing a three-dimensional vehicle model disclosed in this embodiment may include:

[0095] S301. Obtain candidate images associated with the target license plate information.

[0096] S302. Perform target detection on the candidate acquired images to determine the vehicle type and color of the candidate vehicles.

[0097] The vehicle type indicates the model of the candidate vehicle, such as sedan, SUV (Sport Utility Vehicle), MPV (Multi-Purpose Vehicle), or van, etc. The vehicle color indicates the exterior color of the candidate vehicle.

[0098] In one implementation, a deep learning model is used to perform target detection on the candidate acquired images to determine the vehicle type and color of the candidate vehicles in the candidate acquired images.

[0099] This disclosure determines the vehicle type and color of candidate vehicles by performing target detection on candidate acquired images, laying a data foundation for subsequent determination of target vehicles based on vehicle type and color.

[0100] S303. Based on the number of candidate vehicles belonging to the vehicle type and vehicle color, select the target vehicle that matches the target license plate information from the candidate vehicles.

[0101] In one implementation, the number of candidate vehicles belonging to each vehicle type and color is counted, and the target vehicle matching the target license plate information is selected from the candidate vehicles based on the vehicle type and color with the largest number of candidate vehicles.

[0102] This disclosure selects the target vehicle that matches the target license plate information from the candidate vehicles based on the number of candidate vehicles belonging to the vehicle type and vehicle color. This achieves the effect of determining the target vehicle based on the distribution statistics of vehicle type and vehicle color, avoiding problems such as cloned vehicles and false captures, and ensuring the accuracy of determining the target vehicle.

[0103] Optionally, S303 includes:

[0104] The vehicle type with the most candidate vehicles is selected as the target vehicle type, and the color of the vehicle with the most candidate vehicles is selected as the target vehicle color. Candidate vehicles belonging to both the target vehicle type and the target vehicle color are selected as the target vehicles.

[0105] For example, suppose there are 20 candidate vehicles belonging to vehicle type A, 15 candidate vehicles belonging to vehicle type B, and 18 candidate vehicles belonging to vehicle type C. If there are 15 candidate vehicles belonging to vehicle color A, 20 candidate vehicles belonging to vehicle color B, and 18 candidate vehicles belonging to vehicle color C, then vehicle type A is selected as the target vehicle type, and vehicle color B is selected as the target vehicle color. Furthermore, candidate vehicles that simultaneously satisfy both vehicle type A and vehicle color B are selected as the target vehicles.

[0106] By identifying the vehicle type with the most candidate vehicles as the target vehicle type and the color with the most candidate vehicles as the target vehicle color, and then identifying candidate vehicles that share both the target vehicle type and color as the target vehicle, the probability of a candidate vehicle satisfying both the target vehicle type and color being the target vehicle is maximized. This approach minimizes the risk of cloned vehicles and false captures, ensuring the accuracy of target vehicle identification.

[0107] S304. Take the candidate acquisition image of the target vehicle as the target acquisition image, and construct a three-dimensional model of the target vehicle based on the target acquisition image.

[0108] In this embodiment, S201-S206 and S301-S303 can be executed independently, that is, the target vehicle can be determined based on vehicle size, or based on vehicle type and vehicle color. S202-S206 and S303-S303 can also be executed in combination:

[0109] S3001. Determine the target acquisition camera based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image.

[0110] S3002. Based on the camera parameters of the target acquisition camera, convert the candidate acquisition images into three-dimensional acquisition images.

[0111] S3003. Based on the 3D acquired images, determine the vehicle attitude angles and vehicle dimensions of the candidate vehicles in the world coordinate system.

[0112] S3004. Determine the average size based on the vehicle size of the candidate vehicles, and determine at least one size range based on the average size and at least one size threshold.

[0113] S3005. Based on the number of vehicle sizes included in the size range, select the first target vehicle that matches the target license plate information from the candidate vehicles.

[0114] S3006. Perform target detection on the first target vehicle to determine its vehicle type and color.

[0115] S3007. Based on the number of first target vehicles belonging to the vehicle type and color, select a second target vehicle from the first target vehicles that matches the target license plate information.

[0116] The specific implementation methods of each step S3001 to S3007 can be found in the descriptions in the above embodiments, and will not be repeated here.

[0117] Figure 4 This is a flowchart of some three-dimensional model verification methods disclosed in the embodiments of this disclosure. It is further optimized and extended based on the above technical solutions and can be combined with the above optional implementation methods.

[0118] like Figure 4 As shown, the three-dimensional model verification method disclosed in this embodiment may include:

[0119] S401. Determine the target verification image from the target acquisition image and obtain the model image when the 3D model is at the target attitude angle; wherein, the target attitude angle is the vehicle attitude angle of the target vehicle in the target verification image.

[0120] In one implementation, at least one image is selected from the target acquisition images as a target verification image, and the vehicle attitude angle of the target vehicle in the target verification image is determined. Then, the 3D model is rotated according to this vehicle attitude angle until the 3D model is at the target attitude angle, where the target attitude angle is the same as the vehicle attitude angle. The 3D model at the target attitude angle is then cropped to obtain a model image, and a scaling method is used to adjust the model image to the same size as the target verification image.

[0121] S402. Perform target segmentation on the target verification image and the model image respectively, and determine the verification mask area and verification key point position of the target vehicle in the target verification image, as well as the target mask area and target key point position of the target vehicle in the model image.

[0122] In one implementation, a target segmentation algorithm is used to segment the target verification image to determine the verification mask region and verification key point locations of the target vehicle in the target verification image. Furthermore, the target segmentation algorithm is used to segment the model image to determine the target mask region and target key point locations of the target vehicle in the model image.

[0123] S403. Verify the 3D model based on the verification mask area, verification key point positions, target mask area, and target key point positions.

[0124] In one implementation, the area of ​​the verification mask region is compared with that of the target mask region, and the position of the verification key point is compared with that of the target key point. The 3D model is then verified based on the comparison results.

[0125] By determining the target verification image from the target acquisition image and obtaining the model image of the 3D model when it is at the target attitude angle, where the target attitude angle is the vehicle attitude angle of the target vehicle in the target verification image, the target verification image and the model image are segmented separately to determine the verification mask area and verification key point position of the target vehicle in the target verification image, as well as the target mask area and target key point position of the target vehicle in the model image. Then, based on the verification mask area, verification key point position, target mask area and target key point position, the 3D model is verified, which achieves the effect of verifying the quality of the 3D model and ensuring the accuracy of the 3D model.

[0126] Optionally, S403 includes:

[0127] S4031. If the area of ​​the verification mask region and the target mask region are different, or the location of the verification key point is different from the location of the target key point, the 3D model is determined to be unqualified.

[0128] In one implementation, the area of ​​the verification mask region of the target vehicle in each target verification image is compared with the target mask region in the corresponding model image, and the location of the verification key point of the target vehicle in each target verification image is compared with the location of the target key point in the corresponding model image. If the number of target verification images with different area or different location is greater than a first number threshold, the three-dimensional model is considered unqualified.

[0129] For example, assuming there are 10 target verification images and the first quantity threshold is 2 images, if the number of target verification images with different area or different position is 3, then the 3D model is determined to be unqualified.

[0130] S4032. If the area of ​​the verification mask region and the target mask region are the same, and the positions of the verification key points and the target key points are the same, the 3D model is deemed qualified.

[0131] In one implementation, the area of ​​the verification mask region of the target vehicle in each target verification image is compared with the target mask region in the corresponding model image, and the location of the verification key point of the target vehicle in each target verification image is compared with the location of the target key point in the corresponding model image. If the number of target verification images with the same area and the same location is greater than a second quantity threshold, the three-dimensional model is considered to be qualified.

[0132] For example, assuming there are 10 target verification images and the second quantity threshold is 8 images, if the number of target verification images with the same area and the same position is 9, then the 3D model is determined to be qualified.

[0133] By identifying unqualified 3D models when the areas of the verification mask region and the target mask region are different, or when the positions of the verification key points and the target key points are different, the accuracy of the 3D models is guaranteed.

[0134] Based on the above embodiment, after S4031, the following steps are also included:

[0135] S404. Based on the vehicle attitude angle of the target vehicle in the target acquisition image and the second attitude angle spacing, determine the second filter image from the target acquisition image; wherein the second attitude angle spacing is less than the first attitude angle spacing.

[0136] In one implementation, if the 3D model is unqualified, the first attitude angle spacing is reduced to the second attitude angle spacing, and the target acquisition image is re-screened to determine the second screened image based on the second attitude angle spacing and the vehicle attitude angle of the target vehicle in the target acquisition image.

[0137] S405. Perform target segmentation on the second filtered image to determine the second mask region and the second key point position of the target vehicle in the second filtered image.

[0138] The second mask region refers to the occlusion area covering the target vehicle in the second filtered image, which can be used to extract the region of interest of the target vehicle. The second keypoint location refers to the location of the target vehicle's feature points in the second filtered image, including but not limited to the tire position, rearview mirror position, door handle position, and sunroof position, etc.

[0139] S406. Based on the second selected image, the second mask area, and the position of the second key point, construct a three-dimensional model of the target vehicle in the second selected image to generate a three-dimensional model.

[0140] In one implementation, the second selected image, the second mask region, and the second key point position are used as inputs to the SFM algorithm, and the algorithm is executed to generate a three-dimensional model of the target vehicle.

[0141] After generating the 3D model, re-execute S401 to S403 to verify the new 3D model until it is determined to be qualified.

[0142] After determining that the 3D model is unqualified, a second screening image is determined from the target acquisition image based on the vehicle attitude angle and the second attitude angle spacing of the target vehicle in the target acquisition image. The second attitude angle spacing is smaller than the first attitude angle spacing. Target segmentation is performed on the second screening image to determine the second mask region and the second key point position of the target vehicle in the second screening image. Based on the second screening image, the second mask region, and the second key point position, a 3D model of the target vehicle in the second screening image is constructed to generate a 3D model. Since the second attitude angle spacing is smaller than the first attitude angle spacing, the number of second screening images is greater than the number of first screening images. This makes the accuracy of the 3D model built based on the second screening images greater than that built based on the first screening images, thus increasing the probability that the 3D model built based on the second screening images is qualified.

[0143] Figure 5 This is a schematic diagram of a vehicle 3D model construction apparatus disclosed in some embodiments of the present disclosure, which can be applied to the situation of performing 3D modeling of vehicles in acquired images. The apparatus of this embodiment can be implemented in software and / or hardware and can be integrated into any electronic device with computing capabilities.

[0144] like Figure 5 As shown, the vehicle 3D model construction device 50 disclosed in this embodiment may include a vehicle attribute information determination module 51, a target acquisition image determination module 52, and a model construction module 53, wherein:

[0145] The vehicle attribute information determination module 51 is used to acquire candidate images associated with the target license plate information and determine the vehicle attribute information of the candidate vehicles in the candidate images.

[0146] The target acquisition image determination module 52 is used to select the target vehicle that matches the target license plate information from the candidate vehicles according to the vehicle attribute information, and use the candidate acquisition image of the target vehicle as the target acquisition image;

[0147] The model building module 53 is used to build a three-dimensional model of the target vehicle based on the target acquired image.

[0148] Optionally, the vehicle attribute information determination module 51 is specifically used for:

[0149] The target acquisition camera is determined based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image;

[0150] Based on the camera parameters of the target acquisition camera, the candidate acquisition images are converted into three-dimensional acquisition images;

[0151] Based on the 3D acquired images, determine the vehicle attitude angles and vehicle dimensions of the candidate vehicles in the world coordinate system.

[0152] Optionally, the vehicle attribute information determination module 51 is specifically used for:

[0153] Target detection is performed on the candidate images to determine the vehicle type and color of the candidate vehicles.

[0154] Optionally, the target acquisition image determination module 52 is specifically used for:

[0155] Based on the vehicle dimensions of the candidate vehicles, determine the average size, and based on the average size and at least one size threshold, determine at least one size range;

[0156] Based on the number of vehicle sizes included in the size range, select the target vehicle from the candidate vehicles that matches the target license plate information.

[0157] Optionally, the target acquisition image determination module 52 is further used for:

[0158] The size range containing the largest number of vehicle sizes is taken as the target size range, and the vehicle sizes contained in the target size range are taken as the target vehicle sizes;

[0159] Select candidate vehicles that match the target vehicle size as the target vehicle.

[0160] Optionally, the target acquisition image determination module 52 is specifically used for:

[0161] Based on the number of candidate vehicles belonging to the vehicle type and color, select the target vehicle that matches the target license plate information from the candidate vehicles.

[0162] Optionally, the target acquisition image determination module 52 is further used for:

[0163] The vehicle type with the most candidate vehicles is selected as the target vehicle type, and the color of the vehicle with the most candidate vehicles is selected as the target vehicle color.

[0164] Candidate vehicles that share both the target vehicle type and the target vehicle color are selected as target vehicles.

[0165] Optional, model building module 53, specifically used for:

[0166] Based on the vehicle attitude angle of the target vehicle in the target acquisition image and the first attitude angle interval, a first filter image is determined from the target acquisition image;

[0167] Target segmentation is performed on the first screened image to determine the first mask region and the first key point location of the target vehicle in the first screened image;

[0168] Based on the first selected image, the first mask region, and the location of the first key point, a three-dimensional model of the target vehicle in the first selected image is constructed to generate a three-dimensional model.

[0169] Optionally, the device also includes a model validation module, specifically used for:

[0170] The target verification image is determined from the target acquisition image, and the model image of the 3D model when it is at the target attitude angle is obtained; where the target attitude angle is the vehicle attitude angle of the target vehicle in the target verification image;

[0171] Target segmentation is performed on the target verification image and the model image respectively to determine the verification mask region and verification key point location of the target vehicle in the target verification image, and the target mask region and target key point location of the target vehicle in the model image.

[0172] The 3D model is validated based on the validation mask area, validation key point location, target mask area, and target key point location.

[0173] Optionally, the model validation module is also used for:

[0174] If the areas of the verification mask region and the target mask region are different, or if the locations of the verification key points and the target key points are different, the 3D model is determined to be unqualified.

[0175] Optionally, the device also includes a model reconstruction module, specifically used for:

[0176] A second filter image is determined from the target acquisition image based on the vehicle attitude angle of the target vehicle in the target acquisition image and the second attitude angle spacing; wherein the second attitude angle spacing is smaller than the first attitude angle spacing.

[0177] Target segmentation is performed on the second filtered image to determine the second mask region and the location of the second key point of the target vehicle in the second filtered image;

[0178] Based on the second selected image, the second mask area, and the location of the second key point, a three-dimensional model of the target vehicle in the second selected image is constructed to generate a three-dimensional model.

[0179] The vehicle 3D model construction apparatus 50 disclosed in this embodiment can execute the vehicle 3D model construction method disclosed in this embodiment, and has the corresponding functional modules and beneficial effects of the method execution. Content not described in detail in this embodiment can be referred to the description in the method embodiments of this disclosure.

[0180] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0181] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0182] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0183] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0184] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0185] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the method for constructing a three-dimensional vehicle model. For example, in some embodiments, the method for constructing a three-dimensional vehicle model can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the method for constructing a three-dimensional vehicle model described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the method for constructing a three-dimensional vehicle model by any other suitable means (e.g., by means of firmware).

[0186] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0187] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0188] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0189] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0190] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0191] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0192] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0193] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for constructing a three-dimensional vehicle model, comprising: Acquire candidate images associated with the target license plate information; wherein, the target license plate information refers to the license plate information that matches the vehicle for which a 3D model needs to be constructed; The target acquisition camera is determined based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image; Based on the camera parameters of the target acquisition camera, the candidate acquisition image is converted into a three-dimensional acquisition image; Target detection is performed on the acquired 3D image to determine the 3D detection bounding box of the candidate vehicle in the acquired 3D image; Based on the angles between the three-dimensional detection frame and the X-axis, Y-axis, and Z-axis in the world coordinate system, the vehicle attitude angle of the candidate vehicle in the world coordinate system is determined. Based on the length value of the detection frame and the scaling ratio of the 3D detection frame, the length value of the candidate vehicle in the world coordinate system is determined; based on the width value of the detection frame and the scaling ratio of the 3D detection frame, the width value of the candidate vehicle in the world coordinate system is determined; based on the height value of the detection frame and the scaling ratio of the 3D detection frame, the height value of the candidate vehicle in the world coordinate system is determined. The vehicle dimensions of the candidate vehicle are determined based on its length, width, and height values ​​in the world coordinate system. Calculate the average size based on the vehicle dimensions of each candidate vehicle; Starting from the average size, and with at least one preset size threshold as the step size, the system expands in two directions, one smaller than the average size and the other larger than the average size, to form at least one size interval until the size interval covers the smallest and largest vehicle size among the candidate vehicles. The size range containing the largest number of vehicle sizes is taken as the target size range, and the vehicle sizes contained in the target size range are taken as the target vehicle sizes; The candidate vehicles of the target vehicle size are selected as the target vehicles that match the target license plate information; The candidate images of the target vehicle are used as the target images. A three-dimensional model of the target vehicle is constructed based on the target image.

2. The method according to claim 1, wherein, Determining the vehicle attribute information of candidate vehicles in the candidate acquired images includes: Target detection is performed on the candidate acquired images to determine the vehicle type and color of the candidate vehicles.

3. The method according to claim 2, wherein, The step of selecting a target vehicle that matches the target license plate information from the candidate vehicles based on the vehicle attribute information includes: Based on the number of candidate vehicles belonging to the vehicle type and color, a target vehicle matching the target license plate information is selected from the candidate vehicles.

4. The method according to claim 3, wherein, The step of selecting a target vehicle matching the target license plate information from the candidate vehicles based on the number of candidate vehicles of vehicle type and vehicle color includes: The vehicle type with the most candidate vehicles is selected as the target vehicle type, and the color of the vehicle with the most candidate vehicles is selected as the target vehicle color. The candidate vehicles that both the target vehicle type and the target vehicle color belong to are selected as the target vehicles.

5. The method according to claim 1, wherein, The step of constructing a 3D model of the target vehicle based on the target acquired image includes: Based on the vehicle attitude angle of the target vehicle in the target acquisition image and the first attitude angle interval, a first filter image is determined from the target acquisition image; The first filtered image is segmented to determine the first mask region and the first key point location of the target vehicle in the first filtered image. Based on the first selected image, the first mask region, and the location of the first key point, a three-dimensional model of the target vehicle in the first selected image is constructed to generate a three-dimensional model.

6. The method according to claim 5, further comprising, after constructing a three-dimensional model of the target vehicle in the first screened image and generating the three-dimensional model: A target verification image is determined from the target acquisition image, and a model image of the three-dimensional model when it is at the target attitude angle is obtained; wherein, the target attitude angle is the vehicle attitude angle of the target vehicle in the target verification image; Target segmentation is performed on the target verification image and the model image respectively to determine the verification mask region and verification key point position of the target vehicle in the target verification image, and the target mask region and target key point position of the target vehicle in the model image; The 3D model is validated based on the validation mask area, the validation key point positions, the target mask area, and the target key point positions.

7. The method according to claim 6, wherein, The step of verifying the 3D model based on the verification mask area, the verification key point positions, the target mask area, and the target key point positions includes: If the areas of the verification mask region and the target mask region are different, or if the positions of the verification key points and the target key points are different, the 3D model is determined to be unqualified.

8. The method according to claim 7, further comprising, after determining that the three-dimensional model is unqualified: A second filter image is determined from the target acquisition image based on the vehicle attitude angle of the target vehicle in the target acquisition image and the second attitude angle spacing; wherein the second attitude angle spacing is smaller than the first attitude angle spacing; Target segmentation is performed on the second filtered image to determine the second mask region and the location of the second key point of the target vehicle in the second filtered image; Based on the second filtered image, the second mask area, and the second key point position, a three-dimensional model of the target vehicle in the second filtered image is constructed to generate a three-dimensional model.

9. A device for constructing a three-dimensional vehicle model, comprising: The vehicle attribute information determination module is used to acquire candidate images associated with the target license plate information; wherein, the target license plate information refers to the license plate information that matches the vehicle for which a 3D model needs to be constructed; The target acquisition camera is determined based on the camera number associated with the candidate acquisition image; wherein, the target acquisition camera is used to acquire the candidate acquisition image; Based on the camera parameters of the target acquisition camera, the candidate acquisition image is converted into a three-dimensional acquisition image; Target detection is performed on the acquired 3D image to determine the 3D detection bounding box of the candidate vehicle in the acquired 3D image; Based on the angles between the three-dimensional detection frame and the X-axis, Y-axis, and Z-axis in the world coordinate system, the vehicle attitude angle of the candidate vehicle in the world coordinate system is determined. Based on the length value of the detection frame and the scaling ratio of the 3D detection frame, the length value of the candidate vehicle in the world coordinate system is determined; based on the width value of the detection frame and the scaling ratio of the 3D detection frame, the width value of the candidate vehicle in the world coordinate system is determined; based on the height value of the detection frame and the scaling ratio of the 3D detection frame, the height value of the candidate vehicle in the world coordinate system is determined. The vehicle dimensions of the candidate vehicle are determined based on its length, width, and height values ​​in the world coordinate system. The target acquisition image determination module is used to calculate the average size based on the vehicle size of each candidate vehicle; Starting from the average size, and with at least one preset size threshold as the step size, the system expands in two directions, one smaller than the average size and the other larger than the average size, to form at least one size interval until the size interval covers the smallest and largest vehicle size among the candidate vehicles. The size range containing the largest number of vehicle sizes is taken as the target size range, and the vehicle sizes contained in the target size range are taken as the target vehicle sizes; The candidate vehicles of the target vehicle size are selected as the target vehicles that match the target license plate information; The candidate images of the target vehicle are used as the target images. The model building module is used to build a three-dimensional model of the target vehicle based on the target acquired image.

10. The apparatus according to claim 9, wherein, The vehicle attribute information determination module is specifically used for: Target detection is performed on the candidate acquired images to determine the vehicle type and color of the candidate vehicles.

11. The apparatus according to claim 10, wherein, The target image acquisition and determination module is specifically used for: Based on the number of candidate vehicles belonging to the vehicle type and color, a target vehicle matching the target license plate information is selected from the candidate vehicles.

12. The apparatus according to claim 11, wherein, The target image acquisition and determination module is further used for: The vehicle type with the most candidate vehicles is selected as the target vehicle type, and the color of the vehicle with the most candidate vehicles is selected as the target vehicle color. The candidate vehicles that both the target vehicle type and the target vehicle color belong to are selected as the target vehicles.

13. The apparatus according to claim 9, wherein, The model building module is specifically used for: Based on the vehicle attitude angle of the target vehicle in the target acquisition image and the first attitude angle interval, a first filter image is determined from the target acquisition image; The first filtered image is segmented to determine the first mask region and the first key point location of the target vehicle in the first filtered image. Based on the first selected image, the first mask region, and the location of the first key point, a three-dimensional model of the target vehicle in the first selected image is constructed to generate a three-dimensional model.

14. The apparatus according to claim 13, wherein, The device also includes a model verification module, specifically used for: A target verification image is determined from the target acquisition image, and a model image of the three-dimensional model when it is at the target attitude angle is obtained; wherein, the target attitude angle is the vehicle attitude angle of the target vehicle in the target verification image; Target segmentation is performed on the target verification image and the model image respectively to determine the verification mask region and verification key point position of the target vehicle in the target verification image, and the target mask region and target key point position of the target vehicle in the model image; The 3D model is validated based on the validation mask area, the validation key point positions, the target mask area, and the target key point positions.

15. The apparatus according to claim 14, wherein, The model validation module is further used for: If the areas of the verification mask region and the target mask region are different, or if the positions of the verification key points and the target key points are different, the 3D model is determined to be unqualified.

16. The apparatus according to claim 15, wherein, The device also includes a model reconstruction module, specifically used for: A second filter image is determined from the target acquisition image based on the vehicle attitude angle of the target vehicle in the target acquisition image and the second attitude angle spacing; wherein the second attitude angle spacing is smaller than the first attitude angle spacing; Target segmentation is performed on the second filtered image to determine the second mask region and the location of the second key point of the target vehicle in the second filtered image; Based on the second filtered image, the second mask area, and the second key point position, a three-dimensional model of the target vehicle in the second filtered image is constructed to generate a three-dimensional model.

17. An electronic device comprising: At least one processor; as well as A memory that is communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein, Computer instructions are used to cause a computer to perform the method according to any one of claims 1-8.

19. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-8.

Images