A fixed camera based vehicle recognition method and system

Abstract

This invention discloses a vehicle recognition method and system based on fixed cameras. The method includes: acquiring vehicle images through multiple fixed cameras; obtaining vehicle attributes through image recognition and acquiring detailed vehicle driving records; constructing a normal driving time matrix among the cameras to reflect the normal driving time of a vehicle between any two cameras; identifying vehicle trips based on the normal driving time matrix; and statistically analyzing the attributes of vehicles in the same trip based on the trip identification results, with the attribute with the highest percentage being the final vehicle identification result. This invention significantly improves the accuracy of vehicle recognition and provides effective support for intelligent vehicle monitoring and management.

Description

Technical Field

[0001] This invention relates to the field of vehicle recognition, and in particular to a vehicle recognition method and system based on a fixed camera. Background Technology

[0002] With the continuous development of Intelligent Transportation Systems (ITS), road cameras have become one of the important tools for acquiring traffic information. Fixed cameras can capture dynamic information of vehicles, including license plate numbers, vehicle body features, and driving trajectories. However, the raw data captured by current cameras lacks a unified analysis and recognition mechanism, making it difficult to accurately statistically analyze vehicle attributes and operating status. Especially in scenarios such as logistics monitoring, vehicle classification, and dispatching, existing technologies have the following shortcomings:

[0003] 1. Isolated analysis of vehicle information: Data from different cameras lacks correlation, making it impossible to form a unified recognition result.

[0004] 2. Low accuracy of attribute recognition: Existing methods have difficulty in reliably recognizing vehicle attributes, especially in complex environments or with poor image quality, resulting in insufficient reliability of the recognition results.

[0005] 3. Lack of global analysis capabilities: It is impossible to perform statistical analysis based on the information captured by multiple cameras, making it difficult to derive reliable vehicle attributes from multiple dimensions. Summary of the Invention

[0006] This invention provides a vehicle recognition method and system based on a fixed camera, aiming to solve the problem of insufficient vehicle information analysis in the prior art. Through the method and system of this invention, the accuracy of vehicle recognition can be significantly improved, providing effective support for intelligent vehicle monitoring and management.

[0007] The technical solution to achieve the objective of this invention is as follows: a vehicle recognition method based on fixed cameras, which improves the accuracy of vehicle recognition and the robustness of the system by constructing a driving time matrix between cameras and an artificial intelligence model, specifically including the following steps:

[0008] Vehicle images are captured by multiple fixed cameras, vehicle attributes are obtained through image recognition, and vehicle driving record details are retrieved.

[0009] Construct a normal driving time matrix between cameras to reflect the normal driving time of a vehicle between any two cameras;

[0010] Vehicle number identification is performed based on the normal driving time matrix between cameras;

[0011] Based on the train number identification results, the vehicle attributes of the same train number are statistically analyzed, and the attribute identification result with the highest proportion is the final vehicle identification result.

[0012] Furthermore, the driving record details include the license plate number, camera ID, and camera capture time.

[0013] Furthermore, vehicle attributes include vehicle type, vehicle size, body color, whether it carries cargo, and the type of cargo carried.

[0014] Furthermore, the normal driving time matrix between the cameras is as follows:

[0015]

[0016] Among them, a ij Let be the normal travel time between the i-th camera and the j-th camera, where 1 ≤ i, j ≤ n, and n is the number of cameras.

[0017] Furthermore, the normal travel time a between the i-th camera and the j-th camera ij The average value method and the median method were used for calculation.

[0018] Furthermore, the driving time of all vehicles from the i-th camera to the j-th camera within a set time range is obtained from the vehicle driving record details. The average value method is used to take the average of all driving times as the normal driving time, and the median method is used to take the median of all driving times as the normal driving time.

[0019] Furthermore, vehicle identification specifically includes: if the same vehicle continues to drive normally without stopping, it is considered the same vehicle; if a stop occurs between two adjacent cameras, the stop before and after the stop are considered different vehicle numbers.

[0020] Furthermore, the method for determining a stay is as follows:

[0021] Based on the normal driving time matrix between cameras, the normal driving time range is determined to be [a ij (1-p),a ij [1+p], if a vehicle's travel time is within the normal travel time range, then the vehicle is determined to have not stopped; if the actual travel time exceeds the upper limit of the normal travel time range, then the vehicle has stopped en route, where p is a set coefficient, a ij Let be the normal travel time between the i-th camera and the j-th camera.

[0022] Furthermore, the range of coefficient p is set as follows: 10% ≤ p ≤ 30%.

[0023] A vehicle recognition system based on a fixed camera includes:

[0024] The camera capture and acquisition module captures vehicle images through multiple fixed cameras and obtains detailed vehicle driving records.

[0025] The vehicle recognition module obtains vehicle attributes through image recognition, constructs a normal driving time matrix between cameras to reflect the normal driving time of a vehicle between any two cameras, performs vehicle number recognition based on the normal driving time matrix between cameras, and statistically analyzes the vehicle attributes of the same vehicle based on the vehicle number recognition results. The attribute recognition result with the highest proportion is the final vehicle recognition result.

[0026] Compared with the prior art, the beneficial results of the present invention are as follows:

[0027] (1) Constructing a driving matrix can obtain the normal driving time range of a vehicle during its journey, which can be used for abnormal driving detection, congestion analysis, and vehicle path tracking, etc.

[0028] (2) Train number identification: Based on whether there is a stop during the journey, a single journey is divided into multiple train numbers; based on the principle that "the vehicle attributes remain unchanged in the same train number", the vehicle attributes in the same train number can be statistically analyzed; the changes in vehicle attributes in different train numbers can be analyzed to accurately identify vehicle behavior;

[0029] (3) By combining image recognition, vehicle navigation matrix, vehicle number recognition, and statistical analysis techniques, the accuracy of vehicle recognition results is improved: image recognition yields the model recognition result for each image; vehicle numbers are identified by forming a matrix and using a vehicle number recognition algorithm; and the model recognition results are statistically analyzed based on the vehicle numbers to obtain the final vehicle recognition result. Compared to the model recognition result, the final result has higher accuracy. Attached Figure Description

[0030] Figure 1 This is a flowchart of the vehicle recognition process.

[0031] Figure 2 Schematic diagram of a vehicle recognition system. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] This invention provides a vehicle recognition method and system based on a fixed camera, aiming to achieve the following functions:

[0034] (1) Capture vehicle information and store it as historical driving record details;

[0035] (2) Construct a driving time matrix between cameras to accurately estimate vehicle trajectories;

[0036] (3) Using artificial intelligence models, identify vehicle attributes and perform statistical analysis by vehicle number to generate reliable vehicle attribute results.

[0037] Combination Figure 1 This embodiment provides a vehicle recognition method based on a fixed camera, which includes the following steps:

[0038] Step 1: The camera captures images of the vehicle.

[0039] Based on fixed road cameras, images of moving vehicles are captured, along with information such as capture time, camera ID, and license plate number. The collected data is uploaded to a server in a standardized format for subsequent analysis and processing.

[0040] Step 2, Vehicle Driving Record Details

[0041] The captured data (including time, image, and camera location) of each vehicle is stored in the database to form a detailed historical driving record, including license plate number, camera ID, camera capture time, etc.

[0042] Step 3: The deep learning model identifies the captured image and obtains the vehicle model recognition result.

[0043] By using a deep learning model to analyze captured vehicle images, key vehicle attributes are identified, including: vehicle type (such as small cars, large trucks, etc.), vehicle size, body color, whether it is carrying cargo, and the type of cargo, to obtain the vehicle model recognition results.

[0044] Step 4, Driving time matrix between cameras

[0045] A camera-based vehicle travel time matrix (hereinafter referred to as the "travel matrix") is constructed. The travel matrix reflects the normal travel time of a vehicle between any two cameras. Normal travel time refers to the time during which the vehicle travels without stopping due to external or internal interference. Based on the travel time matrix, it can be determined whether a vehicle stopped between two cameras. For example, if a dump truck stops to dump waste between two cameras during its journey, comparing the actual travel time between these two cameras with the normal travel time can determine whether the dump truck stopped between the cameras.

[0046] Assuming there are n cameras, the driving matrix is ​​composed of n×n numbers a ij The resulting n-order square matrix is ​​shown below:

[0047]

[0048] Among them, a ij The normal driving time between the i-th camera and the j-th camera can be calculated using methods such as the average method or the median method. For example, if we obtain the driving time of all vehicles between the i-th camera and the j-th camera within a certain time range (e.g., the last 180 days) from the vehicle's driving record details, if we use the average method, we take the average of all driving times as the normal driving time; if we use the median method, we take the median of all driving times as the normal driving time.

[0049] In reality, due to factors such as road conditions and speed differences, the vehicle's travel time between the two cameras may fluctuate. To accommodate these fluctuations, the normal travel time range is set to [a ij (1-p),a ij [1+p], where 0≤p≤100%, and generally 10%≤p≤30%, for example: p=20%. If the travel time of a vehicle is within the normal travel time range, it can be determined that the vehicle did not stop; if the actual travel time is greater than the upper limit of the normal travel time range, the vehicle stopped on the way, for example: stopping to dump construction waste, or causing traffic congestion.

[0050] Step 5, Vehicle Number Identification

[0051] If the same vehicle (based on its license plate number) travels continuously without stopping, it is considered the same trip. Assume a vehicle travels from its starting point, passing no cameras between the first and m-th cameras, but stopping between the m-th and m+1-th cameras. The journey from the first camera to the m-th camera is counted as one trip. The next trip will begin at the m+1-th camera and end at the camera it passes before its next stop.

[0052] Generally, the attributes of a vehicle remain unchanged within the same trip, such as "whether it is carrying cargo" and "the type of cargo carried".

[0053] Step 6: Count the vehicles by trip number to obtain the final vehicle identification results.

[0054] In step 3, a deep learning model is used to identify a single image, resulting in a vehicle model recognition result. However, the recognition result based on a single image can have significant errors due to factors such as camera resolution, angle, and weather conditions, leading to different recognition results for different captured images of the same vehicle.

[0055] Therefore, in this invention, by utilizing the principle that "vehicle attributes remain unchanged within the same train journey," we can obtain more accurate and reliable recognition results by statistically analyzing the model recognition results within the same train journey. When there are multiple image recognition results for the same attribute of the same vehicle, the recognition result with the highest proportion is the final recognition result for that attribute; the final recognition result for each attribute is statistically calculated to obtain the final recognition result for that vehicle.

[0056] For example, if vehicle A passes by three cameras on the same trip, resulting in three captured images, and the model identification results for two images indicate "yes" to whether vehicle A is carrying cargo, while the model identification result for one image indicates "no," then based on the statistical results, it is more credible to conclude that the vehicle is carrying cargo. Therefore, compared to the identification results from a single camera, vehicle identification based on trip statistics can improve the accuracy and reliability of vehicle identification results.

[0057] Combination Figure 2 This embodiment also provides a vehicle recognition system based on a fixed camera, including:

[0058] 1. Camera capture and data acquisition module, including:

[0059] • Camera Information Management: Manage camera information, including camera name, location, type, camera angle, etc.

[0060] • Camera capture settings: Supports setting the camera capture time, number of captures, etc.

[0061] • Vehicle capture images: Capture images of passing vehicles according to the camera capture settings.

[0062] • Data capture: First, capture data is collected and stored, including license plate number, capture time, camera data, etc., to form historical driving record data; second, the captured images are stored and transmitted to the vehicle recognition module.

[0063] 2. Vehicle recognition module, including:

[0064] • Vehicle recognition: The deep learning model identifies vehicle attributes from the received vehicle images, such as vehicle type (e.g., small car, large truck), vehicle size, body color, whether it is carrying cargo, and the type of cargo.

[0065] • Vehicle number identification: First, construct a driving time matrix based on the vehicle driving record details, and set an update cycle (e.g., daily, weekly, etc.) to update the data of the driving matrix; Second, calculate the vehicle number: construct a vehicle number identification algorithm based on the principle that "the same vehicle (based on the license plate number) is considered to be the same vehicle if it drives continuously and normally without stopping".

[0066] • Vehicle recognition statistics by vehicle: Statistically analyze the attribute results of vehicles by vehicle and output the final recognition result of the vehicle: When there are multiple image recognition results for the same attribute of the same vehicle, the recognition result with the highest proportion is the final recognition result of that attribute; Statistically calculate the final recognition result of each attribute to obtain the final recognition result of the vehicle.

[0067] • Data services: Provides a data service interface that allows third parties to access vehicle recognition results.

[0068] This invention utilizes fixed roadside cameras to capture images of vehicles and combines key technologies such as image recognition, vehicle navigation matrix, and vehicle number recognition to accurately identify vehicle attribute information. The method and system of this invention can significantly improve the accuracy of vehicle identification, providing an efficient and reliable solution for traffic management, logistics monitoring, and smart city construction.

[0069] The description of this invention is given for illustrative and descriptive purposes only and is not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A vehicle recognition method based on a fixed camera, characterized in that, Including the following steps: Vehicle images are captured by multiple fixed cameras, vehicle attributes are obtained through image recognition, and vehicle driving record details are retrieved. Construct a normal driving time matrix between cameras to reflect the normal driving time of a vehicle between any two cameras; Vehicle number identification is performed based on the normal driving time matrix between cameras; Based on the train number identification results, the vehicle attributes of the same train number are statistically analyzed, and the attribute with the highest proportion is the final vehicle identification result. The normal driving time matrix between the cameras is as follows: Among them, a ij Let be the normal travel time between the i-th camera and the j-th camera, where 1 ≤ i, j ≤ n, and n is the number of cameras; Vehicle identification specifically includes: if the same vehicle continues to travel normally without stopping, it is considered the same vehicle; if a stop occurs between two adjacent cameras, the period before and after the stop is considered a different vehicle. The method for determining a stop is as follows: Based on the normal driving time matrix between cameras, the normal driving time range is determined to be [a ij *(1-p),a ij *(1+p)], if a vehicle's travel time is within the normal travel time range, then the vehicle is determined to have not stopped; if the actual travel time exceeds the upper limit of the normal travel time range, then the vehicle has stopped en route, where p is a set coefficient, a ij Let be the normal travel time between the i-th camera and the j-th camera.

2. The vehicle recognition method based on a fixed camera according to claim 1, characterized in that, The driving record details include the license plate number, camera ID, and camera capture time.

3. The vehicle recognition method based on a fixed camera according to claim 1, characterized in that, Vehicle attributes include vehicle type, vehicle size, body color, whether it carries cargo, and the type of cargo carried.

4. The vehicle recognition method based on a fixed camera according to claim 1, characterized in that, The normal travel time a between the i-th camera and the j-th camera ij The average value method and the median method were used for calculation.

5. A vehicle recognition method based on a fixed camera according to claim 4, characterized in that, The driving time of all vehicles from the i-th camera to the j-th camera within a set time range is obtained from the vehicle driving record details. The average value method is used to take the average of all driving times as the normal driving time. The median method is used to take the median of all driving times as the normal driving time.

6. The vehicle recognition method based on a fixed camera according to claim 1, characterized in that, The range of the coefficient p is set as follows: 10% ≤ p ≤ 30%.

7. A vehicle identification system implementing the method of any one of claims 1-6, characterized in that, include: The camera capture and acquisition module captures vehicle images through multiple fixed cameras and obtains detailed vehicle driving records. The vehicle recognition module obtains vehicle attributes through image recognition, constructs a normal driving time matrix between cameras to reflect the normal driving time of a vehicle between any two cameras, performs vehicle number recognition based on the normal driving time matrix between cameras, and statistically analyzes the vehicle attributes of the same vehicle based on the vehicle number recognition results. The attribute recognition result with the highest proportion is the final vehicle recognition result.

Images