Vehicle type determination method, apparatus, processor, and electronic device

Abstract

This invention provides a method, apparatus, processor, and electronic device for determining vehicle type. The method further includes: acquiring three-dimensional point cloud data of a target vehicle, wherein the three-dimensional point cloud data is used to characterize the outline of the target vehicle; converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle; determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one pre-stored standard feature array corresponding to at least one standard vehicle model, wherein the standard feature array is used to characterize the classification result of the standard vehicle model; and determining the classification result corresponding to the target feature array as the target classification result for the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle. This invention solves the technical problem of high resource consumption in the vehicle recognition process.

Description

Technical Field

[0001] The present invention relates to the field of vehicle recognition technology, and more specifically, to a method, apparatus, processor, and electronic device for determining vehicle type. Background Technology

[0002] Currently, roadside lidar is widely used in various scenarios in the transportation field, such as vehicle identification at highway entrances and exits, and traffic flow surveys.

[0003] In related technologies, vehicle type identification is usually achieved by using deep neural network models to identify data collected by roadside LiDAR to obtain vehicle type identification results. However, this method requires a large amount of computation and has the technical problem of high resource consumption during the vehicle identification process.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method, apparatus, processor, and electronic device for determining vehicle type, in order to at least solve the technical problem of high resource consumption during vehicle identification.

[0006] According to an embodiment of the present invention, a method for determining a vehicle type is provided, comprising: acquiring three-dimensional point cloud data of a target vehicle, wherein the three-dimensional point cloud data is used to characterize the outline of the target vehicle; converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle; determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one standard feature array corresponding to at least one standard vehicle model that is pre-stored, wherein the standard feature array is used to characterize the classification result of the standard vehicle model; and determining the classification result corresponding to the target feature array as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

[0007] In an exemplary embodiment, determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one standard feature array corresponding to at least one pre-stored standard vehicle model includes: determining the classification criteria of the target vehicle, wherein the classification results corresponding to different classification criteria are different; and determining the target feature array from at least one standard feature array based on the classification criteria.

[0008] In one exemplary embodiment, determining the target feature array based on the classification criterion from at least one of the standard feature arrays includes: determining error data between the one-dimensional feature array and the standard feature array in response to the classification criterion being the size data of the target vehicle; determining the distance between the one-dimensional feature array and the standard feature array based on the error data, wherein the distance is associated with the similarity; and determining the target feature array based on the distance.

[0009] In an exemplary embodiment, determining the target feature array based on the classification criterion from at least one of the standard feature arrays includes: in response to the classification criterion being the outline shape of the target vehicle, transforming the standard feature array to obtain a first rectangular coordinate corresponding to the standard feature array, and transforming the one-dimensional feature array to obtain a second rectangular coordinate corresponding to the one-dimensional feature array; performing dynamic time warping matching on the first rectangular coordinate and the second rectangular coordinate to obtain an initial matching result; normalizing the initial matching result to obtain a target matching result; and determining the target feature vector based on the target matching result.

[0010] In one exemplary embodiment, the method further includes: determining the body length or body height of the target vehicle; based on the body length or body height, determining at least one standard vehicle model to be matched with the target vehicle from at least one standard vehicle model; and obtaining a standard feature array of the standard vehicle model to be matched.

[0011] In one exemplary embodiment, the three-dimensional point cloud data includes at least one of the following: a first sub-three-dimensional point cloud data corresponding to the side view outline of the target vehicle, a second sub-three-dimensional point cloud data corresponding to the rear view outline of the target vehicle, and a third sub-three-dimensional point cloud data corresponding to the top view outline of the target vehicle.

[0012] In an exemplary embodiment, the step of converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle includes: converting the first sub-three-dimensional point cloud data to obtain a first sub-feature array; converting the second sub-three-dimensional point cloud data to obtain a second sub-feature array; converting the third sub-three-dimensional point cloud data to obtain a third sub-feature array; and combining the first sub-feature array, the second sub-feature array, and the third sub-feature array to obtain the one-dimensional feature array.

[0013] In one exemplary embodiment, acquiring three-dimensional point cloud data of a target vehicle includes: scanning the target vehicle using a multi-line lidar and / or a single-line lidar to obtain three-dimensional point cloud data.

[0014] According to another embodiment of the present invention, a vehicle type determination device is provided, comprising: an acquisition unit for acquiring three-dimensional point cloud data of a target vehicle, wherein the three-dimensional point cloud data is used to characterize the side profile of the target vehicle; a conversion unit for converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle; a first determination unit for determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one standard feature array corresponding to at least one standard vehicle model that is pre-stored, wherein the standard feature array is used to characterize the classification result of the standard vehicle model; and a second determination unit for determining the classification result corresponding to the target feature array as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

[0015] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium storing a plurality of instructions adapted for loading by a processor and executing the steps in any of the above method embodiments.

[0016] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0017] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0018] Vehicle type identification typically utilizes deep neural network models, but this method requires a large amount of data for model training and data processing, resulting in high resource consumption. This invention addresses this issue by pre-acquiring at least one standard feature vector corresponding to at least one standard vehicle model. After acquiring 3D point cloud data, it can be converted into a one-dimensional feature array. This one-dimensional feature array can intuitively describe the complex external features of the target vehicle. By comparing the similarity between the standard feature array and the one-dimensional feature array, a standard vehicle model with high similarity to the target vehicle can be identified from at least one standard vehicle model. Since the classification result corresponding to the standard vehicle model has been pre-determined, this classification result can be used as the target classification result for the target vehicle, thus quickly determining the vehicle type. This solves the problem of high resource consumption in vehicle identification and achieves a reduction in resource consumption during the identification process. Attached Figure Description

[0019] Figure 1 This is a hardware structure block diagram of a mobile terminal for a method of determining vehicle type according to an embodiment of the present invention.

[0020] Figure 2 This is a flowchart of a method for determining a vehicle type according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the side profile of a vehicle according to an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of a coordinate system according to an embodiment of the present invention;

[0023] Figure 5 This is a structural block diagram of a vehicle type determination device according to an embodiment of the present invention;

[0024] Figure 6 This is a schematic diagram of an electronic device for a method of determining vehicle type according to an embodiment of the present invention. Detailed Implementation

[0025] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0027] The methods and embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a vehicle type determination method according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0028] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the vehicle type determination method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the aforementioned method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0029] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0030] Those skilled in the art will understand that Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0031] In one possible implementation, embodiments of this disclosure provide a method for determining vehicle type. Figure 2 This is a flowchart of a method for determining a vehicle type according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0032] Step S202: Obtain the three-dimensional point cloud data of the target vehicle, wherein the three-dimensional point cloud data is used to characterize the outline of the target vehicle.

[0033] In the technical solution provided in step S202 of the present invention, the aforementioned three-dimensional point cloud data can be raw three-dimensional point cloud data, or three-dimensional point cloud information, including a side three-dimensional point cloud map of the vehicle, a top-view three-dimensional point cloud map, etc., which can be used to characterize the side profile of the target vehicle. The aforementioned profile can include side-view profile, rear-view profile, and top-view profile, etc. It should be noted that this is only an example and there is no specific limitation on the data type of the three-dimensional point cloud data.

[0034] In this embodiment, a roadside multi-line lidar can be used to dynamically scan moving vehicles and continuously collect point cloud frame data. A point cloud registration algorithm is then used to fuse point cloud data from different time points to generate a complete and continuous 3D point cloud map of the vehicle, thus obtaining the 3D point cloud data of the target vehicle.

[0035] Optionally, the 3D point cloud data can be composed of X, Y, and Z coordinates. The X-axis represents the vehicle's direction of travel, the Y-axis is perpendicular to the road surface indicating the vehicle's width, and the Z-axis is perpendicular to the ground indicating the vehicle's height. Each point in the 3D point cloud data can represent a reflection point on the vehicle's side at different locations. Multiple 3D point cloud datasets can be used to depict the contour details of the vehicle's side profile, such as changes in height, cabin structure, and window positions. By processing the aforementioned 3D point cloud data, the vehicle's length, width, height, and other dimensional information, as well as side shape features, can be extracted for subsequent vehicle model classification.

[0036] For example, 3D point cloud data can be obtained by combining point clouds from multiple single frames during the vehicle's movement using point cloud registration technology. Due to the varying relative positions of the vehicle and the roadside LiDAR, a single frame of point cloud data may be incomplete at the front, rear, or sides of the vehicle. Point cloud registration technology can stitch together point clouds of the same vehicle from different moments, providing more detailed and accurate vehicle outline information. It should be noted that this is merely an example and does not impose specific limitations on the method of acquiring 3D point cloud data.

[0037] It should be noted that, in addition to processing 3D point cloud data, 1D point cloud data can also be processed, and the following steps can be performed to obtain the target classification result corresponding to the target vehicle.

[0038] Step S204: The three-dimensional point cloud data is converted to obtain a one-dimensional feature array corresponding to the target vehicle.

[0039] In the technical solution provided in step S204 of the present invention, the aforementioned one-dimensional feature array can be a one-dimensional feature vector, also known as a one-dimensional vector. The aforementioned transformation can include projection, coordinate transformation, sampling, filtering, interpolation, dimensionality reduction, etc. It should be noted that this is merely an illustrative example, and no specific limitations are imposed on the transformation method. The aforementioned one-dimensional feature array can be a feature vector.

[0040] In this embodiment, a series of techniques such as projection, coordinate transformation, sampling, filtering, and interpolation can be used to reduce the dimensionality of 3D point cloud data into a fixed-dimensional one-dimensional feature vector. It should be noted that the method of obtaining the one-dimensional feature array described here is merely illustrative and not intended to impose specific limitations. Any method of obtaining a one-dimensional feature array based on 3D point cloud data should be within the scope of protection of this application.

[0041] Optionally, at least one-dimensional feature array corresponding to the target vehicle can be obtained by performing dimensionality reduction processing on the three-dimensional point cloud data.

[0042] Optionally, the three-dimensional point cloud data of the target vehicle can be acquired, and the three-dimensional point cloud data can be transformed to obtain a one-dimensional vector of the target vehicle in multiple dimensions.

[0043] Step S206: In at least one standard feature array corresponding to at least one standard vehicle model that is stored in advance, a target feature array that has a similarity higher than the similarity threshold with the one-dimensional feature array is determined, wherein the standard feature array is used to characterize the classification result of the standard vehicle model.

[0044] In the technical solution provided in step S206 of the present invention, the aforementioned standard feature array can be standard data corresponding to a standard vehicle model, which can be used to characterize the contour information corresponding to the standard vehicle model. The aforementioned similarity threshold can be a preset value, and there are no specific restrictions on the method of determining the similarity threshold here.

[0045] In this embodiment, standard feature data (i.e., standard feature array) corresponding to standard vehicles is obtained in advance. There is a correspondence between the standard feature data and the classification results. That is, the classification criteria corresponding to the one-dimensional feature array can be determined by using the standard feature array.

[0046] Optionally, the original 3D point cloud data can be reduced to a fixed-dimensional one-dimensional feature array through a series of techniques such as projection, coordinate transformation, sampling, filtering, and interpolation. This transforms the point cloud classification problem into an evaluation problem of the distance between feature arrays. Based on existing standard feature arrays of multiple standard vehicle models, the category corresponding to the standard data closest to the undetermined vector can be used as the classification result. In other words, a target feature array with a similarity higher than the similarity threshold to the one-dimensional feature array can be found in at least one standard feature array. Using this target feature array, the side profile of each standard vehicle in the standard library that is most "similar" to the current vehicle's profile can be determined. Based on this standard vehicle model, the classification result of the target vehicle can be determined.

[0047] Optionally, points in the 3D point cloud data correspond one-to-one with one-dimensional feature arrays, and one-dimensional feature data corresponds one-to-one with standard feature arrays. A standard vehicle can correspond to multiple standard feature arrays.

[0048] Step S208: The classification result corresponding to the target feature array is determined as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

[0049] In the technical solution provided by step S208 of the present invention, the classification results can be used to determine the vehicle type corresponding to the standard vehicle. The vehicle type can include types corresponding to different classification standards, such as cargo vans, flatbed vans, container vans, box trucks, tank trucks, concrete mixer trucks, garbage trucks, truck-mounted cranes, etc. It can also include vehicle types determined by a Class I traffic classification standard and vehicle types determined by a Class II traffic classification standard. It should be noted that the classification results here are only illustrative and are not specifically limited.

[0050] In this embodiment, in at least one standard feature array, a target feature array with a similarity higher than a similarity threshold with the one-dimensional feature array is determined, so that the classification result of the standard vehicle corresponding to the target feature array can be determined as the target classification result of the target vehicle.

[0051] For example, a roadside lidar can be used to scan a moving van (i.e., the target vehicle) to collect point cloud frames. Through point cloud registration and data fusion, a continuous and complete 3D side view point cloud image is generated, thus obtaining the 3D point cloud data of the target vehicle. The 3D point cloud data can be transformed to obtain a one-dimensional feature array corresponding to the target vehicle. In a standard library (where the standard feature arrays of standard vehicle models are known), the similarity between the one-dimensional feature array and the standard feature array can be evaluated. Assuming that the one-dimensional feature array of the van has the highest similarity (also called matching degree) with the standard feature array of a certain standard "medium-sized van" in the library, exceeding the similarity threshold T, then the classification result of the target vehicle can be determined as "medium-sized van". Through the above steps, complex 3D point cloud data is transformed into a one-dimensional feature array, and vector feature matching is used to efficiently and accurately complete vehicle type classification.

[0052] Through steps S202 to S208, three-dimensional point cloud data of the target vehicle is obtained, wherein the three-dimensional point cloud data is used to represent the outline of the target vehicle; the three-dimensional point cloud data is transformed to obtain a one-dimensional feature array corresponding to the target vehicle; in at least one standard feature array corresponding to at least one pre-stored standard vehicle model, a target feature array with a similarity higher than the similarity threshold with the one-dimensional feature array is determined, wherein the standard feature array is used to represent the classification result of the standard vehicle model; the classification result corresponding to the target feature array is determined as the target classification result of the target vehicle, wherein the target classification result is used to represent the vehicle type of the target vehicle, thereby solving the technical problem of high resource consumption rate in the vehicle recognition process and achieving the technical effect of reducing resource consumption rate in the vehicle recognition process.

[0053] The embodiments of the present invention will now be described in detail with reference to the steps described above.

[0054] As an optional embodiment, step S206, determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold in at least one standard feature array corresponding to at least one standard vehicle model that is stored in advance, includes: determining the classification standard of the target vehicle, wherein the classification results corresponding to different classification standards are different; and determining the target feature array in at least one standard feature array based on the classification standard.

[0055] In this embodiment, the aforementioned classification criteria can be used to determine the classification scenario, specifically whether it distinguishes vehicle types or classifies vehicle types for traffic jams. Different classification criteria result in different classification outcomes. For example, if the classification criteria only determine the vehicle type, the corresponding classification outcome could include cargo trucks such as stake-on, flatbed, container, and van trucks, as well as tank trucks, concrete mixer trucks, garbage trucks, and truck-mounted cranes. If the classification criteria are a type of traffic jam vehicle classification, used to determine the corresponding type of traffic jam vehicle, the corresponding classification outcome could include automobiles (including medium-sized, large, and small vehicles) and motorcycles. It should be noted that the above classification criteria and results are merely illustrative examples and are not intended to impose specific limitations.

[0056] Optionally, different methods are used to determine the similar target feature array for different classification scenarios. Therefore, before determining the target feature array, the classification criteria of the target vehicle can be determined first. Based on the classification criteria, the target feature array can be determined from at least one standard feature array based on the similarity calculation method corresponding to the classification criteria.

[0057] Optionally, this embodiment can determine the similarity between the standard feature array and the one-dimensional feature array by calculating the distance between them. The following provides further examples of the method for calculating the similarity.

[0058] For example, standard feature arrays corresponding to different standard vehicle models can be stored in advance in a standard vehicle model feature database. These standard feature arrays can be obtained by converting the 3D point cloud data of the standard vehicle models. The aforementioned standard vehicle model feature database can contain a set of one-dimensional vectors (Stand_Veh_Vec) corresponding to at least one standard vehicle model. If the number of known vehicle models is N, and the lengths of the side-view, rear-view, and top-view feature vectors are L1, L2, and L3, respectively, then the final dimension of the above set is N*(L1+L2+L3). Depending on the application scenario of vehicle model classification, the feature vector lengths L1, L2, and L3 can be adjusted or set to 0, representing that the feature vectors from this viewpoint are not considered.

[0059] As an optional embodiment, determining the target feature array based on the classification criteria in at least one of the standard feature arrays includes: determining error data between the one-dimensional feature array and the standard feature array in response to the classification criteria being the size data of the target vehicle; determining the distance between the one-dimensional feature array and the standard feature array based on the error data, wherein the distance is associated with the similarity; and determining the target feature array based on the distance.

[0060] In this embodiment, the aforementioned error data can be the mean square error (MSE) or root mean square error (RMSE) between the standard feature array and the one-dimensional feature array. This error data can be used to characterize the difference between the standard feature array and the one-dimensional feature array, that is, to characterize the similarity between them. The distance is directly proportional to the similarity; that is, the larger the distance, the lower the similarity between the standard feature array and the one-dimensional feature array.

[0061] Optionally, by calculating the MSE or RMSE between two feature arrays (i.e., the standard feature array and the one-dimensional feature array), their average numerical difference can be obtained, thereby determining the degree of similarity between the vehicle to be classified and the standard model. The smaller the MSE or RMSE value, the more similar the two vectors are, and the more likely the vehicles belong to the same category.

[0062] Optionally, in response to the classification criterion of the target vehicle's size data, error data between the one-dimensional feature array and at least one standard feature array can be determined. This error data can be used to characterize the distance between the feature vector and the standard feature array. Based on at least one distance corresponding to the at least one standard feature array, the target feature array with the highest similarity to the feature vector can be determined. Furthermore, based on the classification result corresponding to the target feature array, the classification result of the target vehicle can be determined. The aforementioned size data can be used to determine the length or width of the target vehicle, etc. This is merely an example, and no specific limitation is made on the type of size data.

[0063] In this embodiment, a classification standard corresponding to the current classification environment can also be determined based on the classification environment, thereby determining the distance between the one-dimensional feature array and the standard feature array based on the classification standard.

[0064] For example, taking the classification of a type of cross-traffic vehicle as an example, the external dimensions are an important reference indicator. A 6-meter van and a 12-meter van belong to different vehicle types, so side-view and top-view feature vectors are more important, while rear-view features contribute little. At the same time, a higher sample size is required in the standard library; different sizes of the same type of vehicle must be included in the standard vehicle library. Therefore, when the classification environment is determined to be classifying a type of cross-traffic vehicle, the classification standard can be determined to be based on the size data of the target vehicle. The similarity between the one-dimensional feature array and the standard feature array can then be determined by analyzing the error data between them.

[0065] Optionally, after obtaining the one-dimensional feature vector (i.e., one-dimensional feature array) of the target vehicle, this feature vector can be compared with the standard feature array stored in the standard vehicle feature library in advance to evaluate the similarity between the one-dimensional feature array and the standard feature array, so as to determine the classification result of the target vehicle.

[0066] Optionally, in the 3D point cloud data, one point corresponds to one 1D feature vector, thus obtaining a 1D feature array corresponding to multiple points.

[0067] Optionally, if the classification criterion is the size data of the target vehicle, then each data point (i.e., the standard feature array) in the standard database Stand_Veh_Vec can be traversed to calculate the similarity between the i-th standard feature array and the one-dimensional feature array. When size data needs to be considered, the MSE or RMSE index can be used to evaluate the distance between the two vectors. The calculation formula is as follows:

[0068]

[0069] Here, dist(i) can be used to represent the distance between feature arrays. (Test_Veh_Vec(j)-Stand_Veh_Vec(i,j))2 can be used to represent the error data between the one-dimensional feature array and the standard feature array. L can be used to represent the number of standard feature arrays. Test_Veh_Vec(j) can be used to represent the one-dimensional feature array corresponding to the target vehicle.

[0070] As an optional embodiment, based on the classification criteria, determining the target feature array from at least one of the standard feature arrays includes: in response to the classification criteria being the outline shape of the target vehicle, transforming the standard feature array to obtain a first rectangular coordinate corresponding to the standard feature array, and transforming the one-dimensional feature array to obtain a second rectangular coordinate corresponding to the one-dimensional feature array; performing dynamic time warping matching on the first rectangular coordinate and the second rectangular coordinate to obtain an initial matching result; normalizing the initial matching result to obtain a target matching result; and determining the target feature vector based on the target matching result.

[0071] In this embodiment, the aforementioned outline shape can be used to determine the external profile of the target vehicle, including side view profile, top view profile, or rear view profile, etc. It should be noted that this is only an example and there is no specific limitation on the type of outline shape.

[0072] Alternatively, if the purpose is simply to differentiate vehicle types (e.g., stake-over-the-shelf, flatbed, container, van, tanker, concrete mixer, garbage truck, truck-mounted crane, etc.), side-view and rear-view feature vectors are equally valuable, while top-view vectors contribute less. The assessment of similarity prioritizes "shape" over "size." Therefore, if vehicle outline shape carries higher weight in the assessment, dynamic time warping (DTW) and distance normalization can be used to determine the target feature array from at least one standard feature array.

[0073] For example, if the classification criterion is the external shape of the target vehicle, each standard feature array in Test_Veh_Vec and the standard database Stand_Veh_Vec can be converted to Cartesian coordinates in the following way to obtain the first Cartesian coordinates corresponding to the standard feature array and the second Cartesian coordinates corresponding to the one-dimensional feature array:

[0074]

[0075] Furthermore, dynamic time warping matching is performed on the first and second rectangular coordinates to obtain the initial matching result: [DIST_dtw,IX,IY]=dtw([X j Yj ],[Y jj Y jj ]), where, [X j Y j [Y] can be used to characterize the second rectangular coordinates corresponding to a one-dimensional feature array. jj Y jj ] can be used to characterize the first right-angle coordinate corresponding to the standard feature array.

[0076] Furthermore, the initial matching result can be normalized to obtain the target matching result: dist(i) = DIST_dtw / length(IX), where length(IX) can be used to normalize the initial matching result and can be used to characterize the number of valid points in the match.

[0077] Optionally, at least one target matching result corresponding to a standard feature array is determined, and the target feature array can be determined based on the target matching result. That is, the standard feature array corresponding to the largest target matching result among the at least one target matching result can be determined as the target feature array.

[0078] As an optional embodiment, the method may further include: determining the body length or body height of the target vehicle; based on the body length or body height, determining at least one standard vehicle model to be matched with the target vehicle from at least one standard vehicle model; and obtaining a standard feature array of the standard vehicle model to be matched.

[0079] In this embodiment, in addition to comparing all pre-stored standard feature arrays and feature vectors to determine the target feature array, the body length or body height of the target vehicle can also be determined. Based on the body length or body height, a standard model to be matched with the target vehicle is determined from at least one standard model, so that the standard feature array of the standard model to be matched can be further matched with the one-dimensional feature array.

[0080] Optionally, to improve data processing efficiency, this embodiment does not require calculating the similarity between the multiple one-dimensional feature arrays corresponding to the target vehicle and the samples in the entire standard feature library separately. Further optimization can be made according to actual needs to improve execution efficiency. For example, the standard feature library can be pre-grouped according to vehicle length / vehicle height to obtain multiple sets of undetermined feature vectors Test_Veh_Vec, so that the target vehicle's body length or body height can be compared with some undetermined feature vectors (i.e., standard feature arrays) in the standard library.

[0081] Optionally, after obtaining the similarity between at least one standard feature array and one-dimensional feature array, the standard library sample type with the highest similarity and higher than the set similarity threshold T can be selected as the final classification result;

[0082] FinalVehType=StandType(argmax(dist_selected))

[0083] where dist_selected=dist(dist>T)

[0084] Alternatively, if no standard feature array exists that matches the one-dimensional feature array (i.e., dist_selected is empty), then the vehicle model can be determined to be a new type and can be added to the standard library.

[0085] As an optional embodiment, the three-dimensional point cloud data includes at least one of the following: a first sub-three-dimensional point cloud data corresponding to the side view outline of the target vehicle, a second sub-three-dimensional point cloud data corresponding to the rear view outline of the target vehicle, and a third sub-three-dimensional point cloud data corresponding to the top view outline of the target vehicle.

[0086] In this embodiment, the aforementioned three-dimensional point cloud data may include at least one of the following: a first sub-three-dimensional point cloud data corresponding to the side view outline of the target vehicle, a second sub-three-dimensional point cloud data corresponding to the rear view outline of the target vehicle, and a third sub-three-dimensional point cloud data corresponding to the top view outline of the target vehicle. Therefore, by converting the aforementioned first sub-three-dimensional point cloud data, second sub-three-dimensional point cloud data, and third sub-three-dimensional point cloud data, a feature vector corresponding to the target vehicle can be obtained. This feature vector can be an outline feature vector, which may include: a side view outline feature vector, a rear view outline feature vector, and a top view outline feature vector.

[0087] Optionally, Figure 3 This is a schematic diagram of the side profile of a vehicle according to an embodiment of the present invention, such as... Figure 3 As shown, the three-dimensional point cloud data can include the first sub-three-dimensional point cloud data corresponding to the side view outline of the target vehicle, the second sub-three-dimensional point cloud data corresponding to the rear view outline of the target vehicle, and the third sub-three-dimensional point cloud data corresponding to the top view outline of the target vehicle, which can be used to determine the side view outline, rear view outline and top view outline of the target vehicle. Figure 2 Different patterns can be used to represent the side view, rear view, and top view of different vehicles.

[0088] Optionally, feature vector extraction (dimensionality reduction) can include three types: side-view profile, rear-view profile, and top-view profile. The side-view profile is suitable for situations with distinct side features or a strong dependence on the absolute values ​​of vehicle length and height, such as distinguishing between passenger / freight vehicles, vans / general cargo vehicles, or small and medium-sized trucks within a class of cross-traffic vehicles. The rear-view profile is suitable for distinguishing between different types of trucks or special vehicles (e.g., flatbed trucks, vans, tankers, cement mixers, etc.), or for assisting in the identification of illegally modified vehicles (e.g., raised side panels). The top-view profile requires higher point cloud density, typically necessitating side mounting of the radar, and is suitable for identifying the number of axles, or situations with a strong dependence on vehicle length and width, such as distinguishing between medium and large trucks within a class of cross-traffic vehicles.

[0089] As an optional embodiment, step S204, which involves converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle, includes: converting the first sub-three-dimensional point cloud data to obtain a first sub-feature array; converting the second sub-three-dimensional point cloud data to obtain a second sub-feature array; converting the third sub-three-dimensional point cloud data to obtain a third sub-feature array; and combining the first sub-feature array, the second sub-feature array, and the third sub-feature array to obtain the one-dimensional feature array.

[0090] In this embodiment, the first sub-3D point cloud data, the second sub-3D point cloud data, and the third sub-3D point cloud data can be transformed separately to obtain a first sub-feature array, a second sub-feature array, and a third sub-feature array. These three sub-feature arrays can then be combined to obtain a one-dimensional feature array. The first sub-feature array can be a side-view outline feature vector. The second sub-feature array can be a rear-view outline feature vector. The third sub-feature array can be a top-view outline feature vector.

[0091] Optionally, the conversion of the 3D point cloud data may include converting the first sub-3D point cloud data corresponding to the side view outline, converting the second sub-3D point cloud data corresponding to the rear view outline, and converting the third sub-3D point cloud data corresponding to the top view outline of the target vehicle.

[0092] In this embodiment, a world coordinate system can be agreed upon first, for example, Figure 4 This is a schematic diagram of a coordinate system according to an embodiment of the present invention, such as... Figure 4 As shown, the positive x-axis can be preset as the vehicle's direction of travel, the positive y-axis points to the left of the vehicle, and the positive z-axis is perpendicular to the ground and pointing upwards. Further, the first sub-feature vector can be obtained through the following steps: 3D point cloud data can be loaded, the vehicle's facing direction checked, and if necessary, flipped to ensure the vehicle is facing right. Assume the vehicle's bounding box center coordinates are (X... C ,YC Z C Point clouds below a set height threshold h0 are removed to avoid interference from near-ground targets such as the ground, tires, and rain. Using the vehicle's centerline as a reference, the cloud extends left and right by a set distance w0, retaining only points in the y-direction between h0 and h0. C -w0 to Y C The point cloud within the +w0 range achieves the effect of cutting off the vehicle's side profile; these point clouds will not affect the subsequent calculation of the side profile contour. Assume the X component of the foremost coordinate of the remaining point cloud is X... max The coordinates of the last end are X min The Z component of the topmost coordinate is Z. max The coordinate of the bottom is Z. min Calculate the length and height, where length L = X max -X min Height H = Z max -Z min We can select the center point coordinates as ((X_max+X_min) / 2,h0), and then use all the remaining point clouds (X... i ,Y i Z i Projected onto the XOZ plane (X i Z i ), calculate the polar coordinates (R) relative to the center point coordinates. i ,θ i The calculation method is as follows:

[0093]

[0094] After calculation, 0° <= θ <= 180°, which is close to the positive direction of the X-axis. The smaller θ is, the better.

[0095] Furthermore, the polar coordinates (R) of the transformed residual point cloud can be... i θ i Sort the points j by θ in ascending order, and calculate the point cloud R within each θ degree interval, with a set angle threshold theta. j The maximum value R_max j If there is no point cloud within this interval, R_max j Assign the value Nan. Linear interpolation can be used to fill empty intervals, i.e., R_max. j The value of Nan is shown in the middle. A threshold averaging filter is applied to the R-max sequence to remove jitter in the outline data caused by low point cloud density. The Nan values ​​at the beginning and end of the R-max sequence are padded with constant values, using the closest non-outlier value, to obtain the final 1D feature vector Test_Veh_Vec.

[0096] Optionally, assuming theta = 2°, then 90 intervals will be formed within the range of 0-180 degrees. The final feature vector dimension will be 1*90. The feature vector R_max represents the distance from the outermost contour of the vehicle to the selected center point within each interval of 2 degrees, from 0° at the front to 180° at the rear. It should be noted that the above conversion method is only for illustrative purposes and no specific limitations are imposed here.

[0097] In this embodiment, the second sub-feature vector (i.e., the rear-view outline feature vector) can be determined based on the following steps: loading the 3D point cloud (i.e., the second sub-3D point cloud), checking the vehicle's front-end orientation, and flipping it if necessary to ensure the vehicle's front-end is facing right, assuming the vehicle's bounding box center coordinates are (X... c Y c Z c Point clouds below a set height threshold h0 can be cropped to avoid interference from near-ground targets such as the ground, tires, and rainwater. Furthermore, it can be assumed that the X component of the foremost coordinate of the remaining point cloud is X... max The coordinates of the last end are X min The Z component of the topmost coordinate is Z. max The coordinate of the bottom is Z. min If the vehicle length L = X max -X min If the value is greater than the set threshold, delete items whose X-direction coordinates are between (X... max -X head X max The point cloud within the range allows for the removal of the truck's cab, retaining only the cargo compartment information. The remaining point cloud can be mirrored along the XOZ plane. i Y i Z i )->(X i -Y i Z i To ensure the rearview profile is symmetrical, the center point coordinates can be selected as ((Y_max+Y_min) / 2, h0), and all remaining point clouds (X... i Y i Z i Projected onto the YOZ plane (Y i Z i The polar coordinates (R) relative to the center point can be calculated using the following formula. i θ i ):

[0098]

[0099] After calculation, 0° <= θ <= 180°, which is close to the positive direction of the Y-axis. The larger θ is, the better.

[0100] Optionally, the polar coordinates (R) of the remaining point cloud after transformation can be used. i θ i Sort the points j by θ in ascending order, and calculate the point cloud R within each θ degree interval, with a set angle threshold theta. j The maximum value R_max j If there is no point cloud within this interval, R_max j Assign the value Nan. Linear interpolation can be used to fill empty intervals, i.e., R_max. j The value of Nan is shown in the middle. A threshold averaging filter is applied to the R_max sequence to remove jitter in the outline data caused by low point cloud density. The Nan values ​​at the beginning and end of the R_max sequence are padded with constant values, using the closest non-Nan value, to obtain the final 1D feature vector, i.e., the second sub-feature vector (Test_Veh_Vec).

[0101] In this embodiment, the calculation of the third sub-feature vector (i.e., the top-view outline feature vector) may include the following steps: loading the 3D point cloud (i.e., the third sub-3D point cloud data). The vehicle's orientation can be checked, and if necessary, rotated to ensure the vehicle is facing right. Assume the vehicle's bounding box center coordinates are (X... c Y c Z c Based on the target vehicle's height, height thresholds h1 and h2 can be determined, and only point clouds with heights between h1 and h2 are retained, i.e., only the tire portion is retained.

[0102] Optionally, the remaining point cloud can be mirrored along the XOZ plane, (X i Y i Z i )->(X i -Y i Z i This ensures the top-view outline is symmetrical. Assume the X component of the remaining point cloud's foremost coordinate is X. max The coordinates of the last end are X min The rightmost coordinate Z component is Y max The leftmost coordinate is Y. min All remaining point clouds (X) can be... i Y i Z i Projected onto the YOZ plane (Y i Z i If the center point coordinates are chosen as ((X_max+X_min) / 2, (Y_max+Y_min) / 2), the polar coordinates (R) relative to the center point coordinates can be calculated using the following formula. i θ i ):

[0103]

[0104] After calculation, 0° <= θ <= 360°, and θ = 0° in the positive direction of the X-axis.

[0105] Optionally, the polar coordinates (R) of the remaining point cloud after transformation can be used. i θ i Sort the points by θ in ascending order, and use a set angle threshold theta, with each theta degree representing an interval. Then, calculate the point cloud R within that interval j. j The maximum value R_max j If there is no point cloud within this interval, R_max j Assign the value Nan. Linear interpolation can be used to fill empty intervals, i.e., R_max. j The value in the middle represents the NaN value. A threshold averaging filter is applied to the R_max sequence to remove jitter in the outline data caused by low point cloud density. The NaN values ​​at the beginning and end of the R_max sequence are padded with constant values, using the closest non-Nan value, resulting in the final 1D feature vector Test_Veh_Vec, which is also the third sub-feature vector.

[0106] Optionally, after obtaining the first sub-feature vector, the second sub-feature vector, and the third sub-feature vector, the first sub-feature vector, the second sub-feature vector, and the third sub-feature vector can be simply combined to obtain a one-dimensional feature array corresponding to the target vector.

[0107] Optionally, the 3D point cloud data can be used to represent points within the target vehicle, and can contain multiple 3D coordinates corresponding to multiple points. Each point in the 3D point cloud data corresponds one-to-one with a one-dimensional feature array. That is, 3D point cloud data of the target vehicle can be collected, and the 3D spatial coordinates corresponding to multiple points in the 3D point cloud data can be transformed to obtain a one-dimensional feature array corresponding to multiple points.

[0108] As an optional embodiment, step S202, acquiring the three-dimensional point cloud data of the target vehicle, includes: scanning the target vehicle using a multi-line lidar and / or a single-line lidar to obtain the three-dimensional point cloud data.

[0109] In this embodiment, in addition to using a multi-line LiDAR to scan the target vehicle to obtain 3D point cloud data, a single-line LiDAR can also be used to scan the target vehicle to obtain 3D point cloud data. Alternatively, a combination of multi-line and single-line LiDAR can be used to scan the target vehicle to obtain 3D point cloud data. It should be noted that this is merely an illustrative example and does not impose specific limitations on the method of obtaining 3D point cloud data.

[0110] Optionally, the 3D point cloud mentioned above can be obtained from sources other than multi-line laser registration. Two single-line lasers can also be used to scan along the vehicle length and width directions respectively. Without registration or projection, polar coordinate transformation, downsampling, filtering and other operations can be performed directly to obtain 3D point cloud data.

[0111] In this embodiment of the invention, at least one standard feature array corresponding to at least one standard vehicle model is pre-acquired. After acquiring three-dimensional point cloud data, the three-dimensional point cloud data can be converted into a one-dimensional feature array. This one-dimensional feature array can intuitively describe the complex outline features of the side and front of the target vehicle. By comparing the similarity between the standard feature array and the one-dimensional feature array, a standard vehicle model with a high similarity to the target vehicle can be determined from at least one standard vehicle model. Since the classification result corresponding to the standard vehicle model has been pre-determined, the classification result corresponding to the standard vehicle model can be determined as the target classification result of the target vehicle. This allows for rapid determination of the vehicle type and solves the technical problem of high resource consumption during vehicle recognition, achieving the technical effect of reducing resource consumption during vehicle recognition.

[0112] The following describes in detail another optional implementation method.

[0113] Currently, roadside LiDAR is widely used in various scenarios in the transportation field, such as vehicle identification at highway entrances and exits, and traffic flow surveys. Vehicle classification based on 3D point cloud data is a common application. Common classification standards include traffic survey vehicle classification and toll vehicle classification standards. In addition, customized requirements may arise in certain scenarios, such as the classification / identification of special vehicles like tanker trucks and concrete mixer trucks.

[0114] Table 1 is a classification standard table for a type of vehicle type. As shown in Table 1, the classification results can include automobiles and motorcycles; automobiles can include small cars, medium cars, large cars and extra-large cars; small cars can include small and medium-sized passenger cars and small trucks; medium cars can include large passenger cars and medium-sized trucks; extra-large cars can include extra-large trucks and container trucks. It should be noted that the types of classification results here are only illustrative examples and no specific restrictions are imposed here.

[0115] Table 1 Classification Standards for Type I Traffic Dispatch Vehicles

[0116]

[0117] Table 2 is a classification standard for vehicle types for toll roads. As shown in Table 2, under this classification standard, the corresponding classification results can include: Class I vehicles, Class II vehicles, Class III vehicles, Class IV vehicles, and Class V vehicles.

[0118] Table 2. Vehicle Classification Standards for Toll Roads

[0119] Model Passenger bus (by number of seats) Freight trucks (based on rated load capacity) Class A vehicles ≤7 seats ≤2 tons Category II vehicles 8-19 seats 2 tons to 5 tons (inclusive) Three types of vehicles 20-39 seats 5 to 10 tons (inclusive) Four types of vehicles ≥40 seats 10-15 ton (inclusive) 20-inch container truck Five types of vehicles Container trucks weighing more than 15 tons and 40 inches in diameter

[0120] In related technologies, vehicle outline information (length, width, and height) can be obtained by computing 3D vehicle point cloud data. This information is then combined with information from other sensors within the system (such as cameras, infrared sensors, coils, and millimeter-wave sensors) (e.g., axle count, license plate number, passenger / cargo identification) to determine the classification result of the target vehicle. The entire process relies primarily on other sensors, supplemented by laser point cloud data. However, this method requires the use of other types of sensors, resulting in high costs. Alternatively, vehicle point cloud information can be used, directly or indirectly (e.g., projected into an image), input into a classification neural network (e.g., deep neural network models such as PointNet and MobileNet) to classify the target vehicle. The classification results are then post-processed to obtain the final vehicle model. This process relies primarily on laser point cloud information. However, this method has poor interpretability, requires significant computation, and places certain demands on equipment performance. Furthermore, the model training requires a large amount of sample data to achieve good results. Therefore, this method suffers from high resource consumption during vehicle type recognition.

[0121] To address the aforementioned issues, this invention proposes a vehicle point cloud classification method based on the similarity of vehicle outlines. This method achieves classification accuracy comparable to deep network methods with lower computational resource consumption, resolving the problem of excessive computational resource consumption in deep model-based vehicle classification methods. The method requires a smaller standard library sample size, needing only one sample for each representative vehicle type, and the standard library is expandable. Furthermore, this method offers good interpretability, intuitively using a 1D vector (i.e., feature vector) to describe the complex outline features of the vehicle's side and front, thereby capturing more detailed shape features and achieving more refined classification. For example, it can classify and identify various vehicle types, such as sedans, pickup trucks, mid-size sedans, vans, cargo vans, container trucks, flatbed trucks, tank trucks, concrete mixer trucks, garbage trucks, and truck-mounted cranes. Moreover, this method does not require the use of deep neural network models, which suffer from high resource consumption during vehicle recognition.

[0122] In this embodiment of the invention, at least one standard feature array corresponding to at least one standard vehicle model is pre-acquired. After acquiring three-dimensional point cloud data, the three-dimensional point cloud data can be converted into a one-dimensional feature array. This one-dimensional feature array can intuitively describe the complex outline features of the side and front of the target vehicle. By comparing the similarity between the standard feature array and the one-dimensional feature array, a standard vehicle model with a high similarity to the target vehicle can be determined from at least one standard vehicle model. Since the classification result corresponding to the standard vehicle model has been pre-determined, the classification result corresponding to the standard vehicle model can be determined as the target classification result of the target vehicle. This allows for rapid determination of the vehicle type and solves the technical problem of high resource consumption during vehicle recognition, achieving the technical effect of reducing resource consumption during vehicle recognition.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0124] This embodiment also provides a vehicle type determination device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0125] Figure 5 This is a structural block diagram of a vehicle type determination device according to an embodiment of the present invention, such as... Figure 5 As shown, the device includes: an acquisition unit 502, a conversion unit 504, a first determination unit 506, and a second determination unit 508.

[0126] The acquisition unit 502 is used to acquire three-dimensional point cloud data of the target vehicle, wherein the three-dimensional point cloud data is used to characterize the side profile of the target vehicle.

[0127] The conversion unit 504 is used to convert the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle; the first determination unit is used to determine the target feature array whose similarity to the one-dimensional feature array is higher than the similarity threshold from at least one standard feature array corresponding to at least one standard vehicle model that is stored in advance, wherein the standard feature array is used to characterize the classification result of the standard vehicle model.

[0128] The second determining unit 506 is used to determine the classification result corresponding to the target feature array as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

[0129] The vehicle type determination device provided in this embodiment of the invention acquires three-dimensional point cloud data of a target vehicle through an acquisition unit, wherein the three-dimensional point cloud data is used to represent the side profile of the target vehicle; a conversion unit converts the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle; a first determination unit determines a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one standard feature array corresponding to at least one standard vehicle model that is pre-stored, wherein the standard feature array is used to represent the classification result of the standard vehicle model; a second determination unit determines the classification result corresponding to the target feature array as the target classification result of the target vehicle, wherein the target classification result is used to represent the vehicle type of the target vehicle, thereby solving the technical problem of high resource consumption rate in the vehicle recognition process and realizing the technical problem of reducing the resource consumption rate of determining the vehicle type.

[0130] It should be noted that the above-mentioned units can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above-mentioned units are located in the same processor; or, the above-mentioned units are located in different processors in any combination.

[0131] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0132] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0133] Figure 6 This is a schematic diagram of the structure of an electronic device for a vehicle type determination method according to an embodiment of the present invention, as shown below. Figure 6 As shown, embodiments of the present invention also provide an electronic device 600, including a processor 601 and a memory 602, wherein the memory 602 stores a computer program, and the processor 601 is configured to run the computer program to perform the steps in any of the above method embodiments.

[0134] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0135] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0136] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0137] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining vehicle type, characterized in that, include: Acquire three-dimensional point cloud data of the target vehicle, wherein the three-dimensional point cloud data is used to characterize the outline of the target vehicle; The three-dimensional point cloud data is transformed to obtain a one-dimensional feature array corresponding to the target vehicle; In at least one standard feature array corresponding to at least one standard vehicle model that is stored in advance, a target feature array with a similarity higher than a similarity threshold with the one-dimensional feature array is determined, wherein the standard feature array is used to characterize the classification result of the standard vehicle model; The classification result corresponding to the target feature array is determined as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

2. The method for determining vehicle type according to claim 1, characterized in that, The step of determining a target feature array whose similarity to the one-dimensional feature array is higher than a similarity threshold from at least one pre-stored standard feature array corresponding to at least one standard vehicle model includes: The classification criteria for the target vehicle are determined, wherein different classification criteria correspond to different classification results; Based on the classification criteria, the target feature array is determined from at least one of the standard feature arrays.

3. The method for determining vehicle type according to claim 2, characterized in that, The step of determining the target feature array based on the classification criteria from at least one of the standard feature arrays includes: In response to the classification criterion being the size data of the target vehicle, error data between the one-dimensional feature array and the standard feature array is determined; Based on the error data, the distance between the one-dimensional feature array and the standard feature array is determined, wherein the distance is related to the similarity. The target feature array is determined based on the distance.

4. The method for determining vehicle type according to claim 2, characterized in that, The step of determining the target feature array based on the classification criteria from at least one of the standard feature arrays includes: In response to the classification criterion being the outline shape of the target vehicle, the standard feature array is transformed to obtain the first rectangular coordinate corresponding to the standard feature array, and the one-dimensional feature array is transformed to obtain the second rectangular coordinate corresponding to the one-dimensional feature array; Dynamic time warping matching is performed on the first rectangular coordinate and the second rectangular coordinate to obtain the initial matching result; The initial matching result is normalized to obtain the target matching result; Based on the target matching result, the target feature vector is determined.

5. The method for determining vehicle type according to any one of claims 1 to 4, characterized in that, The method further includes: Determine the body length or body height of the target vehicle; Based on the vehicle length or vehicle height, at least one standard vehicle model to be matched with the target vehicle is determined from at least one of the standard vehicle models. Obtain the standard feature array of the standard vehicle model to be matched.

6. The method for determining vehicle type according to claim 1, characterized in that, The three-dimensional point cloud data includes at least one of the following: a first sub-three-dimensional point cloud data corresponding to the side view outline of the target vehicle, a second sub-three-dimensional point cloud data corresponding to the rear view outline of the target vehicle, and a third sub-three-dimensional point cloud data corresponding to the top view outline of the target vehicle.

7. The method for determining vehicle type according to claim 6, characterized in that, The process of converting the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle includes: The first sub-3D point cloud data is transformed to obtain the first sub-feature array; The second sub-3D point cloud data is transformed to obtain the second sub-feature array; The third sub-3D point cloud data is transformed to obtain the third sub-feature array; The first sub-feature array, the second sub-feature array, and the third sub-feature array are combined to obtain the one-dimensional feature array.

8. The method for determining vehicle type according to claim 1, characterized in that, The acquisition of the target vehicle's 3D point cloud data includes: The target vehicle is scanned using a multi-line lidar and / or a single-line lidar to obtain the three-dimensional point cloud data.

9. A vehicle type determining device, characterized in that, include: An acquisition unit is used to acquire three-dimensional point cloud data of a target vehicle, wherein the three-dimensional point cloud data is used to characterize the side profile of the target vehicle; The conversion unit is used to convert the three-dimensional point cloud data to obtain a one-dimensional feature array corresponding to the target vehicle. The first determining unit is configured to determine, from at least one standard feature array corresponding to at least one standard vehicle model that is stored in advance, a target feature array that has a similarity higher than a similarity threshold with the one-dimensional feature array, wherein the standard feature array is used to characterize the classification result of the standard vehicle model; The second determining unit is used to determine the classification result corresponding to the target feature array as the target classification result of the target vehicle, wherein the target classification result is used to characterize the vehicle type of the target vehicle.

10. A processor, characterized in that, The processor is used to run a program, wherein the program is executed by the processor to perform the method according to any one of claims 1 to 8.

Images