CN115719465B - Vehicle detection method, device, apparatus, storage medium, and program product - Google Patents

Abstract

The disclosure provides a vehicle detection method, a device, equipment, a storage medium and a program product, and relates to the technical field of data processing, in particular to the technical fields of big data, artificial intelligence and intelligent traffic. The specific implementation scheme is as follows: determining a vehicle region according to the input image data, wherein the vehicle region represents a region where a vehicle displayed by the image data is located; and determining vehicle detection data according to the vehicle region, wherein the vehicle detection data comprises at least one of the following: the vehicle comprises a vehicle attribute, a vehicle license number, a violation detection result, a vehicle running track and a vehicle flow detection result, wherein the vehicle attribute represents the physical characteristics of the vehicle.

Description

Vehicle detection method, device, apparatus, storage medium, and program product

Technical Field

The present disclosure relates to the field of data processing technologies, and in particular, to the field of big data, artificial intelligence, and intelligent traffic technologies, and in particular, to a vehicle detection method, apparatus, device, storage medium, and program product.

Background

Vehicles have various demands such as daily management as a common vehicle. With the development of computer technology and internet technology, related technologies can assist in vehicle detection and the like.

Disclosure of Invention

The present disclosure provides a vehicle detection method, apparatus, device, storage medium, and program product.

According to an aspect of the present disclosure, there is provided a vehicle detection method including: determining a vehicle region according to the input image data, wherein the vehicle region represents a region where a vehicle displayed by the image data is located; and determining vehicle detection data according to the vehicle region, wherein the vehicle detection data comprises at least one of the following: the vehicle comprises a vehicle attribute, a vehicle license number, a violation detection result, a vehicle running track and a vehicle flow detection result, wherein the vehicle attribute represents the physical characteristics of the vehicle.

According to another aspect of the present disclosure, there is provided a vehicle detection apparatus including: a vehicle region determining module and a vehicle detection data determining module. And the vehicle region determining module is used for determining a vehicle region according to the input image data, wherein the vehicle region represents the region where the vehicle displayed by the image data is located. A vehicle detection data determining module for determining vehicle detection data according to a vehicle region, wherein the vehicle detection data includes at least one of: the vehicle comprises a vehicle attribute, a vehicle license number, a violation detection result, a vehicle running track and a vehicle flow detection result, wherein the vehicle attribute represents the physical characteristics of the vehicle.

According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor and a memory communicatively coupled to the at least one processor. Wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the methods of the embodiments of the present disclosure.

According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of the embodiments of the present disclosure.

According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program stored on at least one of a readable storage medium and an electronic device, the computer program when executed by a processor implementing a method of an embodiment of the present disclosure.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 schematically illustrates a system architecture diagram of a vehicle detection method and apparatus according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a flow chart of a vehicle detection method according to an embodiment of the disclosure;

FIG. 3 schematically illustrates a schematic diagram of a vehicle detection method according to another embodiment of the present disclosure;

fig. 4 schematically illustrates a schematic diagram of obtaining a vehicle license number according to a vehicle detection method of a further embodiment of the present disclosure;

fig. 5 schematically shows a specific example of displaying a vehicle license number license and a vehicle color as a vehicle attribute in a time-series video frame Img;

FIG. 6 schematically illustrates a block diagram of a vehicle detection apparatus according to an embodiment of the disclosure; and

fig. 7 schematically illustrates a block diagram of an electronic device that may implement a vehicle detection method of an embodiment of the disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

Fig. 1 schematically illustrates a system architecture of a vehicle detection method and apparatus according to an embodiment of the present disclosure. It should be noted that fig. 1 is only an example of a system architecture to which embodiments of the present disclosure may be applied to assist those skilled in the art in understanding the technical content of the present disclosure, but does not mean that embodiments of the present disclosure may not be used in other devices, systems, environments, or scenarios.

As shown in fig. 1, a system architecture 100 according to this embodiment may include clients 101, 102, 103, a network 104, and a server 105. The network 104 is the medium used to provide communication links between the clients 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

A user may interact with the server 105 through the network 104 using clients 101, 102, 103 to receive or send messages, etc. Various communication client applications may be installed on clients 101, 102, 103, such as shopping class applications, web browser applications, search class applications, instant messaging tools, mailbox clients, social platform software, and the like (by way of example only).

The clients 101, 102, 103 may be a variety of electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like. The clients 101, 102, 103 of the disclosed embodiments may, for example, run applications.

The server 105 may be a server providing various services, such as a background management server (by way of example only) that provides support for websites browsed by users using clients 101, 102, 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the client. In addition, the server 105 may also be a cloud server, i.e. the server 105 has cloud computing functionality.

It should be noted that the vehicle detection method provided by the embodiment of the present disclosure may be executed by the server 105. Accordingly, the vehicle detection device provided by the embodiment of the present disclosure may be provided in the server 105. The vehicle detection method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and that is capable of communicating with the clients 101, 102, 103 and/or the server 105. Accordingly, the vehicle detection apparatus provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the clients 101, 102, 103 and/or the server 105.

In one example, the server 105 may acquire input image data from the clients 101, 102, 103 through the network 104, and perform vehicle detection on the image data, determining a vehicle detection result.

It should be understood that the number of clients, networks, and servers in fig. 1 is merely illustrative. There may be any number of clients, networks, and servers, as desired for implementation.

It should be noted that, in the technical solution of the present disclosure, the processes of collecting, storing, using, processing, transmitting, providing, disclosing, etc. related personal information of the user all conform to the rules of the related laws and regulations, and do not violate the public welfare.

In the technical scheme of the disclosure, the authorization or consent of the user is obtained before the personal information of the user is obtained or acquired.

The embodiment of the present disclosure provides a vehicle detection method, and a vehicle detection method according to an exemplary embodiment of the present disclosure is described below with reference to fig. 2 to 5 in conjunction with the system architecture of fig. 1. The vehicle detection method of the embodiment of the present disclosure may be performed by the server 105 shown in fig. 1, for example.

Fig. 2 schematically illustrates a flow chart of a vehicle detection method according to an embodiment of the present disclosure.

As shown in fig. 2, the vehicle detection method 200 of the embodiment of the present disclosure may include, for example, operations S210 to S220.

In operation S210, a vehicle region is determined from the input image data.

The input image data may be received from terminals 101, 102, 103, for example. The image data may be captured by an image capturing device such as a camera of a road, for example.

The vehicle region represents the region where the vehicle displayed by the image data is located.

For example, the vehicle region may be marked with a detection box, for example.

In operation S220, vehicle detection data is determined according to the vehicle region.

The vehicle detection data includes at least one of: vehicle attributes, vehicle license numbers, violation detection results, vehicle running tracks and vehicle flow detection results.

The vehicle attribute characterizes a physical feature of the vehicle.

By way of example, the vehicle attribute may include, for example, a color of the vehicle, a model of the vehicle, and the like.

According to the vehicle detection method, the vehicle region is determined according to the input image data, so that the vehicle region can be positioned in the input image data to display the vehicle region in the image data, and the vehicle detection data can be accurately determined based on the vehicle region conveniently.

According to the vehicle detection method of the embodiment of the present disclosure, the through-vehicle detection data includes at least one of: the vehicle attribute, the vehicle license number, the violation detection result, the vehicle running track and the vehicle flow detection result can realize multifunctional and comprehensive vehicle detection. For example, at least one of the vehicle attribute, the vehicle license number, the violation detection result, the vehicle running track and the vehicle flow detection result can basically cover the vehicle detection requirements in various scenes, and has higher vehicle detection efficiency.

Fig. 3 schematically illustrates a schematic diagram of a vehicle detection method 300 according to another embodiment of the present disclosure.

Image data 301 may include a plurality of time-sequential video frames. The input image data 301 may be, for example, at least one individual image 301-1, or may be a video 301-2 or time-sequential video frames of a real-time video stream 301-2, each time-sequential video frame being in an image format. In the example of fig. 3, a specific example in which the video 301-2 includes a total of N sequential video frames of the sequential video frame P1 to the sequential video frame PN is schematically shown. In the example of fig. 3, a specific example in which the video stream 301-3 includes a total of K sequential video frames of the sequential video frame PF1 to the sequential video frame PFK is also schematically shown.

In the case where the input image data is one, for example, a vehicle attribute or a vehicle license number may be determined. In the case where the input image data is plural or the input image data includes plural time-series video frames, the five types of vehicle detection junctions described above, that is, the vehicle attribute, the vehicle license number, the violation detection result, the vehicle running track, and the traffic flow detection result, may be determined.

As shown in fig. 3, a vehicle detection method 300 according to another embodiment of the present disclosure, with a specific example of determining vehicle detection data according to a vehicle region may be implemented using the following embodiments.

Illustratively, determining the vehicle detection data includes at least one of operations S321-S325, for example, according to the vehicle region. In the example of fig. 3, a specific example is schematically shown in which the determination of the vehicle detection data includes, for example, operations S321 to S325, in accordance with the vehicle region.

In operation S321, the vehicle attribute 303 is determined from the vehicle region 302.

As shown in fig. 3, a specific example of determining the vehicle attribute according to the vehicle region of operation S321 may be exemplarily implemented using, for example, the following embodiments: the vehicle region 302 is detected according to the vehicle attribute detection model M2, and the vehicle attribute 303 is obtained.

By way of example, the vehicle attribute detection model may include, for example: paddlePaddle-an Lightweight CPU Convolutional Neural Network, PP-LCNet, is a lightweight convolutional neural network.

The PP-LCNet can improve the accuracy of vehicle attribute detection under the condition of keeping the delay unchanged. The PP-LCNet is lighter, has higher vehicle attribute detection speed and vehicle attribute detection accuracy, has lower memory requirement, and is suitable for vehicle detection in more scenes.

In operation S322, the vehicle license plate number 305 is determined according to the vehicle region 302.

As shown in fig. 3, a specific example of determining the vehicle license number 305 from the vehicle region 302 of operation S322 may be exemplarily implemented using the following embodiments, for example.

In operation S326, the vehicle region 302 is detected according to the license plate detection model M3, and the license plate region 304 is obtained.

By way of example, the vehicle license plate detection model may include, for example: paddlePaddle-Optical Character Recognition vesion3_det, namely PP-OCRv3_det.

The PP-OCRv3_det has a pixel aggregation network structure with a large receptive field, a teacher model mutual learning strategy and a characteristic pyramid structure of a residual error attention mechanism, and can efficiently and accurately detect license plates of vehicles.

In operation S327, the license plate region 304 is detected according to the character recognition model M4 to obtain the license plate number 305.

By way of example, the character recognition model M4 may include, for example, paddlePaddle-Optical Character Recognition vesion3_rec, namely PP-OCRv3_rec.

PP-OCRv3_rec is a lightweight text recognition network, recognition accuracy of vehicle license plate numbers can be improved by replacing a cyclic neural network RNN with a small-scale high-precision Chinese scene text recognition model SVTR_Tiny, and meanwhile, speed of vehicle license plate number detection can be improved by utilizing a attention-directed CTC training strategy, a data augmentation strategy for excavating text context information, a self-supervised pre-training model, a joint mutual learning strategy and a non-labeling data excavation scheme, so that a lightweight character recognition model M4 with high character recognition accuracy and high speed is provided.

In operation S323, the violation detection result 306 is determined based on the vehicle license number and the vehicle regions of the plurality of time-series video frames.

The violation detection result relates to the vehicle's violation and the vehicle license number associated with the violation. In some cases, the violation of the vehicle needs to be determined according to continuous behavior in the time dimension, and thus the violation detection result is related to a plurality of time-series video frames.

By way of example, a specific example of determining the violation detection result from the vehicle license number and the vehicle region of the plurality of time-sequential video frames may be implemented using the following embodiments: a vehicle identification associated with the vehicle license plate region is determined. And determining an illegal parking detection result of the vehicle according to the vehicle region related to the illegal parking region and the vehicle identifier and the illegal parking time threshold aiming at any one vehicle identifier.

The offending detection results include an offending parking detection result.

The vehicle identifier may be obtained by marking the vehicle region of the input image data, for example, a corresponding vehicle identifier may be marked in each vehicle region of the initial input image data, the vehicle identifiers of the subsequent time-series video frames may be synchronized by vehicle tracking, for example, the same vehicle identifier may be associated, or a vehicle identifier may be added, deleted, or the like.

For example, the out-of-order parking area and the out-of-order parking time threshold may be pre-configured, for example, according to a specific scenario.

It should be noted that the vehicle identifier may be used to uniquely identify the vehicle, and the vehicle identifier may be the same as or different from the vehicle license plate number.

The illegal parking is a common type of illegal parking detection, and the vehicle detection method of the embodiment of the disclosure can cover a conventional illegal parking detection function by detecting the illegal parking.

In operation S324, the vehicle running track 307 is determined from the vehicle regions of the plurality of time-series video frames.

As shown in fig. 3, a specific example of the vehicle running track 307 may be determined, for example, by implementing the following embodiment for the vehicle region of operation S324 for a plurality of time-series video frames: the vehicle region 302 of the plurality of time-series video frames is detected according to the vehicle tracking model M5, and the vehicle running track 307 is obtained.

The vehicle tracking model M5 may include, for example: the hybridization-center SORT: rethinking SORT for Robust Multi-Object Tracking, OC-SORT, a multi-objective Tracking algorithm.

The OC-SORT is suitable for the conditions that the model is sensitive to target overlapping and nonlinear motion and high-frame-rate video is needed in multi-target tracking, can simply track vehicles on line and in real time, and has the characteristic of robustness in the process of target overlapping and nonlinear motion.

In operation S325, the vehicle flow rate detection result 308 is determined based on the vehicle running trajectory 307 and the vehicle region.

By way of example, a specific example of determining the vehicle flow detection result from the vehicle running track and the vehicle region may be implemented using the following embodiments: and determining a vehicle flow detection result in the corresponding direction according to the number of the vehicle areas and the direction of the vehicle running track.

For example, in some scenarios, traffic flow detection needs to be accurate to the direction of vehicle travel. According to the vehicle detection method disclosed by the embodiment of the invention, the vehicle flow detection result in the corresponding direction can be accurately determined through the number of the vehicle areas and the direction of the vehicle running track.

According to the vehicle detection method disclosed by the embodiment of the disclosure, through the operation, for example, the vehicle attribute, the vehicle license number, the violation detection result, the vehicle running track and the vehicle flow detection result can be automatically determined, so that the multifunctional and full-scene vehicle detection is realized.

According to the vehicle detection method, high-dimensional input image data can be handled through the corresponding deep learning model, and corresponding detection can be accurately performed by using the expression capability of deep learning.

According to the vehicle detection method disclosed by the embodiment of the disclosure, by means of a vehicle attribute detection model such as PP-LCNet, a vehicle license plate detection model such as PP-OCRv3_det, a character recognition model such as PP-OCRv3_rec and a vehicle tracking model such as OC-SORT, technical model selection is not required for related users, and the usability is higher.

In some scenes, the data and model magnitude related to the realization of the multifunctional and full-scene vehicle detection all need large-memory and high-performance hardware equipment, and according to the vehicle detection method of the embodiment of the disclosure, the specific deep learning model has the advantages of light weight, high detection accuracy, high detection speed and the like, and can reduce the memory and performance requirements on the hardware equipment and has higher adaptability with the hardware equipment under the condition of realizing the multifunctional and full-scene vehicle detection.

In addition, according to the vehicle detection method of the embodiment of the present disclosure, a pipeline operation may be systematically performed, and at least one type of vehicle detection result may be output for input image data through a corresponding operation. For example, the violation detection result may be determined based on a vehicle detection result of the type of the license plate number of the vehicle, and the traffic flow detection result may be determined based on a vehicle detection result of the type of the vehicle running track. The vehicle detection method and the vehicle detection system can provide various vehicle detection under the condition that the vehicle detection results comprise vehicle attributes, vehicle license plate numbers, violation detection results, vehicle running tracks and vehicle flow detection results, and have higher vehicle detection efficiency without the need of logic construction of corresponding detection functions or adaptive adjustment of corresponding deep learning models by corresponding users.

Fig. 4 schematically illustrates a schematic diagram of obtaining a vehicle license number according to a vehicle detection method of a further embodiment of the present disclosure.

As shown in fig. 4, according to a vehicle detection method according to still another embodiment of the present disclosure, detection of a vehicle license plate region according to a character recognition model may be implemented using the following embodiments, resulting in a specific example of a vehicle license plate number.

In operation S471, the vehicle identification 402 associated with the vehicle license plate region 401 is determined.

Detecting a vehicle license number requires a predetermined vehicle license area and is associated with a vehicle identification, e.g., a vehicle appears in a plurality of time sequential video frames, with which the corresponding vehicle identification is associated, detecting the vehicle license number of the vehicle.

In operation S472, the license plate region of the vehicle with the time-series video frame 403 of the vehicle identification 402 is detected according to the character recognition model, and the candidate license plate number 404 is obtained.

In operation S473, the maximum number of candidate vehicle license numbers having the same numerical value is determined as the vehicle license number 405.

In the following, an example will be described in which a certain vehicle C1 is displayed in a total of X time-series video frames from the time-series video frame P1 to the time-series video frame PX. For example, in some cases, the license plate number of the candidate vehicle detected by some time-series video frames is wrong due to different angles of the vehicle C1 in the different time-series video frames or the presence of a shade or the like.

According to the vehicle detection method of the embodiment of the disclosure, the vehicle license plate area of the time sequence video frame with the vehicle identification can be uniquely determined based on the vehicle identification by determining the vehicle identification related to the vehicle license plate area, the candidate vehicle license plate numbers are obtained by detecting the vehicle license plate area of the time sequence video frame with the vehicle identification according to the character recognition model, and the maximum number of candidate vehicle license plate numbers with the same value are determined as the vehicle license plate numbers, so that the vehicle license plate number detection accuracy can be improved.

In the example of fig. 4, the candidate vehicle license number detected by the vehicle license area of each time series video frame is schematically shown, for example, the candidate vehicle license number N1 corresponding to the time series video frame P1 is 12345, the candidate vehicle license number N3 corresponding to the time series video frame P3 is 12845, and in the case where the number of candidate vehicle license numbers having the value of 12345 is the largest, the candidate vehicle license number 12345 may be determined as the vehicle license number 405.

Illustratively, a vehicle detection method according to still another embodiment of the present disclosure may implement, for example, a specific example of determining vehicle detection data according to a vehicle region using the following embodiments: in response to the vehicle detection data selection instruction, vehicle detection data matching the vehicle detection data selection instruction is determined according to the vehicle region.

For example, the vehicle detection data selection instruction may be determined by a person having a need for vehicle detection, for example.

According to the vehicle detection method, the vehicle detection data are selected, the vehicle detection data matched with the vehicle detection data selection instruction can be determined according to the vehicle region, the vehicle detection data are selected and configured, the flexibility is higher, and the vehicle detection method can be suitable for more application scenes of vehicle detection.

Illustratively, according to the vehicle detection method of the further embodiment of the present disclosure, a specific example of determining the vehicle region from the input image data may be implemented, for example, with the following embodiments: a plurality of first target sequential video frames sequentially spaced apart from M video frames are determined based on the plurality of consecutive sequential video frames. And detecting the target time sequence video frame to obtain a vehicle region.

M is a positive integer greater than 1. By way of example, M may take on values in the range of an integer number from 2 to 12.

In the case where the input image data includes a plurality of time-series video frames, the interval time between the time-series video frames is short, the difference between adjacent time-series video frames is small, and detecting each time-series video frame in turn has a large amount of computation effort, time, storage cost, and the accuracy of vehicle detection is not significantly improved.

According to the vehicle detection method disclosed by the embodiment of the disclosure, the number of time sequence video frames needing to be detected in the vehicle region can be reduced by determining the first target time sequence video frames of a plurality of M video frames which are sequentially spaced according to the plurality of continuous time sequence video frames, and the vehicle region can be obtained by detecting the target time sequence video frames, so that the calculation power consumption can be reduced, the calculation time and the storage capacity can be saved, and the vehicle detection efficiency is higher.

For example, in the case where the vehicle detection data includes a vehicle running track, since the vehicle running track has a higher correlation with time, M may be given a lower value, for example, in order to ensure accuracy of the vehicle running track. M may take on a value of 2, for example.

Illustratively, a vehicle detection method according to still another embodiment of the present disclosure may further include, for example: and responding to a structure adjustment instruction aiming at the target model, and adjusting the structure of the target model to obtain an adjusted target model.

The object model includes at least one of an object detection model, a vehicle attribute detection model, a vehicle license plate detection model, a character recognition model, and a vehicle tracking model.

The deep learning model has a structure such as a head network (head), a backbone network (backbone), or the like.

According to the vehicle detection method, the structure of the target model is selected, the structure of the target model is adjusted in response to the structure adjustment instruction for the target model, the adjusted target model can be obtained, the structure selection and configuration of the target network are supported, the flexibility is higher, and the vehicle detection method can be suitable for more application scenes of vehicle detection.

Illustratively, according to a vehicle detection method of a further embodiment of the present disclosure, the target detection model is trained using a generic training set and a business training set.

The general training set characterizes the training set for target detection and the business training set characterizes the training set for vehicle detection.

The generic training set may be obtained, for example, on an open-source training sample platform.

The object detection model is used for object detection, and the object may include various object objects, and for vehicle region detection according to the embodiments of the present disclosure, model training needs to be adaptively performed, which involves training samples related to vehicles such as road driving. These training samples are costly to the relevant personnel or entities with the need for vehicle detection.

According to the vehicle detection method, the target detection model obtained through training by utilizing the general training set and the business training set can be directly used, the model and the training model are not required to be built, and the technical threshold of related personnel or entities with vehicle detection requirements can be reduced.

According to the vehicle detection method, the target detection model obtained through training of the general training set and the service training set can be directly suitable for application scenes of vehicle detection, related personnel or entities with vehicle detection requirements are not required to acquire the training set related to the vehicle for adaptive model training, cost of vehicle detection can be reduced, and the vehicle detection method disclosed by the embodiment of the invention has higher usability and vehicle detection efficiency.

Illustratively, according to a vehicle detection method of a further embodiment of the present disclosure, the target detection model is obtained by performing model fine-tuning according to a generic training set and a business training set.

Model tuning, i.e., fine-tune, can be understood as a method of tuning a model to bridge such distribution shifts over data that more closely approximates the target data distribution.

Model fine tuning may include, for example: a target detection model, i.e., a source model, is pre-trained in the source training set. A new object detection model is created that replicates all model designs and their parameters on the source model except the output layer. It is assumed that the new target detection model parameters contain knowledge learned on the source training set and that these knowledge apply equally to the target training set. It is also assumed that the output layer of the source model is closely related to the labels of the source training set. An output layer with the output size being the number of the target training set categories can be added for the new target detection model, and model parameters of the layer are initialized randomly. A new target detection model may be trained on the target training set. The output layer may be trained from scratch, while the parameters of the remaining layers are all derived based on fine-tuning of the parameters of the source model.

For example, model fine-tuning may be used where different training sets are applied for target detection model training, for example.

For example, for training sets with small amounts of data but very high data similarity, the output class of the last layers or the final classification layers of the object detection model may be modified.

For example, for a training set with a small amount of data but low similarity, the initial k-layers of the pre-trained target detection model may be frozen and the remaining (n-k) layers are trained there.

According to the vehicle detection method disclosed by the embodiment of the invention, the target detection model obtained by carrying out model fine adjustment according to the general training set and the service training set has better performance, can accurately and efficiently detect the vehicle region, and has higher vehicle detection efficiency.

For example, the object detection model may include: paddlePaddle-an Evolved version of YOLO, PP-yoloE.

Under the condition that the target detection model comprises PP-YOLOE, model training convergence can be accelerated through the PP-YOLOE, and downstream task generalization is improved.

Illustratively, according to a further embodiment of the present disclosure, the vehicle attribute detection model is trained using a data enhancement training set.

The data enhancement training set is obtained by carrying out data enhancement processing on the training set of the vehicle attribute detection model.

Data enhancement can be understood as a method in which existing data generates new training data to extend the original training set.

According to the vehicle detection method of the embodiment of the disclosure, as the data enhancement training set is obtained by carrying out data enhancement processing on the training set of the vehicle attribute detection model, the cost of data acquisition and data marking of the training set of the vehicle attribute detection model is lower, the unbalance of the training set is reduced, the generalization of the vehicle attribute detection model obtained by training the data enhancement training set is better, the accuracy of vehicle attribute detection is higher,

illustratively, according to the vehicle detection method of the further embodiment of the present disclosure, the vehicle attribute detection model is obtained by performing transfer learning on the student model by using the teacher model.

Illustratively, the teacher model is more complex in structure, for example, and the model parameters are more massive, for example, than the student model.

According to the vehicle detection method disclosed by the embodiment of the disclosure, the vehicle attribute detection model obtained by utilizing the teacher model to carry out transfer learning on the student model is lighter, the accuracy of vehicle attribute detection of the teacher model can be achieved, and the vehicle attribute detection efficiency is higher.

Illustratively, according to a vehicle detection method of a further embodiment of the present disclosure, the training set of the vehicle attribute detection model includes a target label sample, which is a labeled training sample obtained by detecting a label-free sample with a teacher model.

One of the more difficult problems with the use of a model of the correlation of a technician or entity with the need for vehicle detection is: how the model applies to the current vertical class scenario. For example, the correlation model is suitable for a wider application scenario, and in a specific application scenario, the performance and detection accuracy of the model can be greatly reduced. The secondary training is needed, and the acquisition and labeling cost of the training sample is high.

According to the vehicle detection method, the training samples with the labels, which are obtained by detecting the unlabeled samples by using the teacher model, can be combined with the teacher model to automatically label the initial unlabeled samples, so that the labor cost of labeling the relevant labels can be saved. Through the training set of the vehicle attribute detection model comprising the target label sample, a more accurate vehicle attribute detection model which is more suitable for the current application scene can be obtained.

For example, the rest of the target models can be obtained by transferring and learning from the teacher model to the student model, and the labeled training samples of the corresponding target models can be labeled by the unlabeled samples through the teacher model.

As shown in fig. 5, the vehicle detection method according to still another embodiment of the present disclosure may further include: the vehicle detection data is displayed in a time-series video frame.

In the example of fig. 5, a specific example of displaying a vehicle license number license and a vehicle color as vehicle attributes in a time-series video frame Img is schematically shown. In the example of fig. 5, a vehicle zone Z1 and a vehicle license plate zone Z2 are also schematically shown.

According to the vehicle detection method of the embodiment of the present disclosure, by displaying the vehicle detection data in the time-series video frame, the vehicle detection data can be visually displayed when the relevant person views the time-series video frame.

Illustratively, the vehicle detection method according to still another embodiment of the present disclosure may also be configured with various parameters and data, such as a save position, an acquisition position, and the like, such as a path configuration supporting input image data, an operation device configuration of the vehicle detection method of the embodiment of the present disclosure, and the like, for example.

Fig. 6 schematically shows a block diagram of a vehicle detection apparatus according to an embodiment of the present disclosure.

As shown in fig. 6, the vehicle detection 600 of the embodiments of the present disclosure includes, for example, a vehicle region determination module 610 and a vehicle detection data determination module 620.

A vehicle region determining module 610, configured to determine a vehicle region according to the input image data, where the vehicle region represents a region where the vehicle displayed by the image data is located; and

the vehicle detection data determining module 620 is configured to determine vehicle detection data according to a vehicle region, where the vehicle detection data includes at least one of: the vehicle comprises a vehicle attribute, a vehicle license number, a violation detection result, a vehicle running track and a vehicle flow detection result, wherein the vehicle attribute represents the physical characteristics of the vehicle.

Illustratively, the image data includes a plurality of time-sequential video frames; the vehicle detection data determination module includes at least one of the following.

The vehicle attribute determination submodule is used for determining the vehicle attribute according to the vehicle area.

And the vehicle license number determining submodule is used for determining the vehicle license number according to the vehicle region.

And the violation detection result determining submodule is used for determining the violation detection result according to the vehicle license number and the vehicle areas of the plurality of time sequence video frames.

And the vehicle running track determining sub-module is used for determining the vehicle running track according to the vehicle areas of the time sequence video frames.

And the traffic flow detection result determining submodule is used for determining a traffic flow detection result according to the vehicle running track and the vehicle area.

Illustratively, the vehicle region determination module includes: and the first vehicle region determining submodule is used for detecting the input image data according to the target detection model to obtain the vehicle region.

The vehicle attribute determination submodule includes: and the vehicle attribute determining unit is used for detecting the vehicle area according to the vehicle attribute detection model to obtain the vehicle attribute.

The license plate number determination submodule includes: the vehicle license plate area determining unit is used for detecting the vehicle area according to the vehicle license plate detection model to obtain a vehicle license plate area; and the vehicle license number determining unit is used for detecting the vehicle license area according to the character recognition model to obtain the vehicle license number.

The violation detection result determining submodule comprises: a vehicle identification determination unit configured to determine a vehicle identification associated with a vehicle license plate region; and the illegal parking detection result determining unit is used for determining the illegal parking detection result of the vehicle according to the vehicle area related to the illegal parking area and the vehicle identifier and the illegal parking time threshold value aiming at any one vehicle identifier, wherein the illegal parking detection result comprises the illegal parking detection result.

The vehicle running track determination submodule includes: and the vehicle running track determining unit is used for detecting the vehicle areas of the time sequence video frames according to the vehicle tracking model to obtain the vehicle running track.

The traffic flow detection result determination submodule includes: and the vehicle flow detection result determining unit is used for determining a vehicle flow detection result in a corresponding direction according to the number of the vehicle areas and the direction of the vehicle running track.

Illustratively, the vehicle license plate number determining unit includes: a vehicle identification determination subunit configured to determine a vehicle identification associated with the license plate region of the vehicle; the candidate vehicle license number determining subunit is used for detecting the vehicle license area of the time sequence video frame with the vehicle identifier according to the character recognition model to obtain a candidate vehicle license number; and a license plate number determining subunit for determining a maximum number of candidate license plate numbers having the same numerical value as the license plate number.

Illustratively, the vehicle region determination module includes: the first target time sequence video frame determining submodule is used for determining a plurality of first target time sequence video frames which are sequentially separated by M video frames according to a plurality of continuous time sequence video frames, wherein M is a positive integer greater than 1; and the second determining submodule of the vehicle region is used for detecting the target time sequence video frame to obtain the vehicle region.

Illustratively, the vehicle detection data determination module includes: and the selection module is used for responding to the vehicle detection data selection instruction and determining vehicle detection data matched with the vehicle detection data selection instruction according to the vehicle area.

Illustratively, the method further comprises: and the adjusting module is used for responding to the structure adjusting instruction aiming at the target model, adjusting the structure of the target model and obtaining an adjusted target model, wherein the target model comprises at least one of a target detection model, a vehicle attribute detection model, a vehicle license plate detection model, a character recognition model and a vehicle tracking model.

The target detection model is illustratively trained using a generic training set that characterizes the training set for target detection and a business training set that characterizes the training set for vehicle detection.

The target detection model is illustratively obtained by performing model fine-tuning according to a general training set and a business training set.

Illustratively, the vehicle attribute detection model is trained using a data enhancement training set obtained by performing data enhancement processing on a training set of the vehicle attribute detection model.

Illustratively, the vehicle attribute detection model is obtained by performing transfer learning on the student model by using a teacher model.

Illustratively, the training set of the vehicle attribute detection model includes a target label sample, which is a labeled training sample obtained by detecting an unlabeled sample with a teacher model.

Illustratively, the object detection model includes: paddlePaddle-an Evolved version of YOLO, vehicle attribute detection models include PaddlePaddle-an Lightweight CPU Convolutional Neural Network, vehicle license plate detection models include PaddlePaddle-Optical Character Recognition vesion3_det, character recognition models include PaddlePaddle-Optical Character Recognition vesion3_rec, and vehicle tracking models include Observation-center SORT: rethinking SORT for Robust Multi-Object Tracking.

Illustratively, the method further comprises: and the display module is used for displaying the vehicle detection data in the time sequence video frames.

It should be understood that the embodiments of the apparatus portion of the present disclosure correspond to the same or similar embodiments of the method portion of the present disclosure, and the technical problems to be solved and the technical effects to be achieved also correspond to the same or similar embodiments, which are not described herein in detail.

According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

Fig. 7 illustrates a schematic block diagram of an example electronic device 700 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 7, the apparatus 700 includes a computing unit 701 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 702 or a computer program loaded from a storage unit 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 may also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other through a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.

Various components in device 700 are connected to I/O interface 705, including: an input unit 706 such as a keyboard, a mouse, etc.; an output unit 707 such as various types of displays, speakers, and the like; a storage unit 708 such as a magnetic disk, an optical disk, or the like; and a communication unit 709 such as a network card, modem, wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.

The computing unit 701 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 701 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized 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 701 performs the respective methods and processes described above, such as a vehicle detection method. For example, in some embodiments, the vehicle detection method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 700 via ROM 702 and/or communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the vehicle detection method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the vehicle detection method by any other suitable means (e.g. by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), complex Programmable Logic Devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These 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 the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

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

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims (26)

1. A vehicle detection method, comprising:

determining a vehicle region according to input image data, wherein the vehicle region represents a region where a vehicle displayed by the image data is located, and the image data comprises a plurality of time sequence video frames;

wherein: the determining a vehicle region from the input image data includes:

detecting the input image data according to a target detection model to obtain the vehicle region;

responding to a structure adjustment instruction aiming at the target model, and adjusting the structure of the target model to obtain an adjusted target model; the target model is obtained by performing model fine adjustment based on a general training set and a business training set; the model fine tuning includes modifying an output class of a model classification layer; the general training set characterization aims at a target detection training set; the service training set characterizes a training set aiming at vehicle detection; the output layer of the target model is closely related to the labels of the general training set; the object model includes: the target detection model, the vehicle attribute detection model, the vehicle license plate detection model, the character recognition model and the vehicle tracking model; wherein, the adjusting the structure of the target model includes: adding output layers with the output size being the number of the target training set categories; and

Determining vehicle detection data according to the vehicle region, wherein the vehicle detection data comprises: the vehicle detection method comprises the steps of vehicle attributes, vehicle license plate numbers, violation detection results, vehicle running tracks and vehicle flow detection results, wherein the vehicle attributes represent physical characteristics of the vehicle;

wherein, according to the vehicle region, determining vehicle detection data includes:

detecting the vehicle region according to a vehicle attribute detection model to obtain the vehicle attribute;

detecting the vehicle region according to a vehicle license plate detection model to obtain a vehicle license plate region;

detecting the license plate area according to a character recognition model to obtain the license plate number of the vehicle;

and detecting the vehicle areas of the time sequence video frames according to a vehicle tracking model to obtain the vehicle running track.

2. The method of claim 1, wherein the determining vehicle detection data from the vehicle region further comprises:

determining the violation detection result according to the vehicle license number and the vehicle areas of a plurality of time sequence video frames; and

and determining the traffic flow detection result according to the vehicle running track and the vehicle area.

3. The method of claim 2, wherein,

the determining the license plate number according to the vehicle region comprises:

said determining said violation detection result from said vehicle license number and said vehicle region of said plurality of time-sequential video frames comprises:

determining a vehicle identification associated with the vehicle license plate region;

determining an illegal parking detection result of the vehicle according to the vehicle region and an illegal parking time threshold associated with the vehicle identifier by aiming at any one vehicle identifier, wherein the illegal parking detection result comprises the illegal parking detection result;

the determining the vehicle trajectory for the vehicle region of the plurality of time-sequential video frames includes:

the determining the traffic flow detection result according to the vehicle running track and the vehicle area comprises the following steps:

and determining a vehicle flow detection result in the corresponding direction according to the number of the vehicle areas and the direction of the vehicle running track.

4. The method of claim 3, wherein the detecting the vehicle license plate region according to the character recognition model, the obtaining the vehicle license plate number comprises:

Determining a vehicle identification associated with the vehicle license plate region;

detecting the license plate area of the time sequence video frame with the vehicle identifier according to the character recognition model to obtain a candidate license plate number;

and determining the maximum number of the candidate license plate numbers with the same numerical value as the license plate numbers.

5. The method of any of claims 2-4, wherein the determining a vehicle region from the input image data comprises:

determining a plurality of first target time sequence video frames which are sequentially separated by M video frames according to a plurality of continuous time sequence video frames, wherein M is a positive integer greater than 1;

and detecting the target time sequence video frame to obtain the vehicle region.

6. The method of any of claims 1-4, wherein the determining vehicle detection data from the vehicle region comprises:

in response to a vehicle detection data selection instruction, the vehicle detection data matching the vehicle detection data selection instruction is determined according to the vehicle region.

7. The method of claim 1, wherein the target detection model is trained using a generic training set that characterizes a training set for target detection and a business training set that characterizes a training set for vehicle detection.

8. The method of claim 1, wherein the vehicle attribute detection model is trained using a data enhancement training set that is a data enhancement processing of a training set of the vehicle attribute detection model.

9. The method according to claim 1, wherein the vehicle attribute detection model is obtained by performing transfer learning on a student model using a teacher model.

10. The method of claim 9, wherein the training set of vehicle attribute detection models includes target label samples, the target label samples being labeled training samples obtained by detecting unlabeled samples using the teacher model.

11. The method of claim 1, wherein the object detection model comprises: paddlePaddle-an Evolved version of YOLO, said vehicle attribute detection model comprises PaddlePaddle-an Lightweight CPU Convolutional Neural Network, said vehicle license plate detection model comprises PaddlePaddle-Optical Character Recognition vesion3_det, said character recognition model comprises PaddlePaddle-Optical Character Recognition vesion3_rec, and said vehicle Tracking model comprises Observation-center SORT: rethinking SORT for Robust Multi-Object Tracking.

12. The method of any of claims 2-4, further comprising:

the vehicle detection data is displayed in the time-series video frame.

13. A vehicle detection apparatus comprising:

the vehicle region determining module is used for determining a vehicle region according to input image data, wherein the vehicle region represents a region where a vehicle displayed by the image data is located, and the image data comprises a plurality of time sequence video frames;

wherein the vehicle region determination module includes:

the first determining submodule of the vehicle region is used for detecting the input image data according to a target detection model to obtain the vehicle region;

the adjusting module is used for responding to a structure adjusting instruction aiming at the target model, adjusting the structure of the target model to obtain an adjusted target model, wherein the target model is obtained by carrying out model fine adjustment based on a general training set and a business training set; the model fine tuning includes modifying an output class of a model classification layer; the general training set characterization aims at a target detection training set; the service training set characterizes a training set aiming at vehicle detection; the output layer of the target model is closely related to the labels of the general training set; wherein the target model includes the target detection model, the vehicle attribute detection model, the vehicle license plate detection model, a character recognition model, and the vehicle tracking model; and

The vehicle detection data determining module is used for determining vehicle detection data according to the vehicle area, wherein the vehicle detection data comprises: the vehicle detection method comprises the steps of vehicle attributes, vehicle license plate numbers, violation detection results, vehicle running tracks and vehicle flow detection results, wherein the vehicle attributes represent physical characteristics of the vehicle;

the vehicle detection data determination module includes:

the vehicle attribute determining unit is used for detecting the vehicle region according to a vehicle attribute detection model to obtain the vehicle attribute;

the vehicle license plate area determining unit is used for detecting the vehicle area according to the vehicle license plate detection model to obtain a vehicle license plate area;

the vehicle license number determining unit is used for detecting the vehicle license area according to the character recognition model to obtain the vehicle license number;

and the vehicle running track determining unit is used for detecting the vehicle areas of the time sequence video frames according to a vehicle tracking model to obtain the vehicle running track.

14. The apparatus of claim 13, wherein the vehicle detection data determination module further comprises:

the violation detection result determining submodule is used for determining the violation detection result according to the vehicle license number and the vehicle areas of the time sequence video frames; and

And the traffic flow detection result determining submodule is used for determining the traffic flow detection result according to the vehicle running track and the vehicle area.

15. The apparatus of claim 14, wherein,

the violation detection result determining submodule comprises:

a vehicle identification determination unit configured to determine a vehicle identification associated with the vehicle license plate region;

the illegal parking detection result determining unit is used for determining an illegal parking detection result of the vehicle according to the vehicle area and the illegal parking time threshold value associated with the illegal parking area and the vehicle identifier, wherein the illegal parking detection result comprises the illegal parking detection result;

the traffic flow detection result determining submodule includes:

and the vehicle flow detection result determining unit is used for determining a vehicle flow detection result in a corresponding direction according to the number of the vehicle areas and the direction of the vehicle running track.

16. The apparatus of claim 15, wherein the vehicle license number determination unit comprises:

a vehicle identification determination subunit configured to determine a vehicle identification associated with the vehicle license plate region;

A candidate vehicle license number determining subunit, configured to detect, according to the character recognition model, the vehicle license area of the time-sequential video frame with the vehicle identifier, to obtain a candidate vehicle license number;

and the vehicle license number determining subunit is used for determining the maximum number of the candidate vehicle license numbers with the same value as the vehicle license number.

17. The apparatus of any of claims 14-16, wherein the vehicle region determination module comprises:

the first target time sequence video frame determining submodule is used for determining a plurality of first target time sequence video frames which are sequentially separated by M video frames according to a plurality of continuous time sequence video frames, wherein M is a positive integer greater than 1;

and the second determining submodule of the vehicle region is used for detecting the target time sequence video frame to obtain the vehicle region.

18. The apparatus of any of claims 13-16, wherein the vehicle detection data determination module comprises:

and the selection module is used for responding to a vehicle detection data selection instruction and determining the vehicle detection data matched with the vehicle detection data selection instruction according to the vehicle area.

19. The apparatus of claim 14, wherein the target detection model is trained using a generic training set that characterizes a training set for target detection and a business training set that characterizes a training set for vehicle detection.

20. The apparatus of claim 14, wherein the vehicle attribute detection model is trained using a data enhancement training set that is a data enhancement processing of a training set of the vehicle attribute detection model.

21. The apparatus of claim 14, wherein the vehicle attribute detection model is a result of a transfer learning of a student model using a teacher model.

22. The apparatus of claim 21, wherein the training set of vehicle attribute detection models includes target label samples, the target label samples being labeled training samples obtained by detecting unlabeled samples using the teacher model.

23. The apparatus of claim 14, wherein the object detection model comprises: paddlePaddle-an Evolved version of YOLO, said vehicle attribute detection model comprises PaddlePaddle-an Lightweight CPU Convolutional NeuralNetwork, said vehicle license plate detection model comprises PaddlePaddle-Optical CharacterRecognition vesion3_det, said character recognition model comprises PaddlePaddle-Optical Character Recognition vesion3_rec, and said vehicle Tracking model comprises Observation-center SORT: rethinking SORT for RobustMulti-Object Tracking.

24. The apparatus of any of claims 14-16, further comprising:

and the display module is used for displaying the vehicle detection data in the time sequence video frame.

25. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-12.

26. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-12.