A vehicle speed measurement method and system based on multi-device fusion

Abstract

The application discloses a kind of vehicle speed determination method and system based on multi-device fusion, belong to traffic engineering technical field, including the construction camera acquisition database and distributed optical fiber acquisition database;Setting detection area, real-time data is obtained in detection area, including the first relative position of the vehicle to be measured, first detection time, second relative position and second detection time;The similarity of first relative position and second relative position is judged, and the target vehicle is determined;Pixel distance and motion distance are calculated;In the corresponding relationship of pixel distance and motion distance in the detection area, the corresponding relationship of pixel coordinates and world coordinates is constructed, and the speed of target vehicle is obtained.The application utilizes the advantage that distributed optical fiber equipment can be accurately positioned, combined with camera equipment, realizes vehicle speed determination based on multi-device fusion, and improves the accuracy of vehicle speed determination.

Description

Technical Field

[0001] This invention relates to the field of traffic engineering technology, and in particular to a method and system for measuring vehicle speed based on multi-device fusion. Background Technology

[0002] Intelligent Transportation Systems (ITS) are comprehensive systems that utilize advanced information and communication technologies to improve traffic management efficiency, enhance traffic safety, and reduce environmental impact. They optimize traffic flow and reduce congestion and accident rates through real-time data collection, analysis, and processing. In the current era of rapid development of ITS, accurate vehicle speed measurement is crucial for optimizing traffic management and ensuring road safety. While significant progress has been made in target detection within the traffic field, numerous challenges remain in vehicle speed measurement.

[0003] Existing patents, such as patent application number CN201210080108.7, disclose a video-based vehicle speed detection method. This method processes acquired traffic video to establish a curve showing the relationship between actual position and time, and then fits the vehicle speed. However, this vehicle speed measurement technology relies on only a single acquisition device, leading to incomplete data and severely impacting the accuracy of speed measurement in complex traffic environments. Furthermore, other existing technologies involve analyzing vehicle movement image sequences captured by cameras installed above the road, processing the images to analyze vehicle displacement, establishing a mapping relationship between pixels and world coordinates, and calculating speed. However, these existing technologies often suffer from inaccurate pixel-to-world coordinate mapping, resulting in unsatisfactory vehicle speed measurements and hindering the provision of reliable data support for traffic management. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a vehicle speed measurement method and system based on multi-device fusion. By leveraging the precise positioning capabilities of distributed fiber optic devices and combining them with camera equipment, it achieves vehicle speed measurement based on multi-device fusion, thereby improving the accuracy of vehicle speed measurement.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a vehicle speed measurement method based on multi-device fusion, comprising:

[0007] Data from the vehicle under test is acquired by cameras and distributed optical fiber, respectively, and corresponding camera acquisition database and distributed optical fiber acquisition database are constructed.

[0008] A detection area is set within the overlapping area of ​​the camera and the distributed optical fiber acquisition. Within the detection area, real-time data of the vehicle under test identified by the camera is acquired, the camera acquisition database is updated, and the first relative position and first detection time of the vehicle under test are generated. Real-time data of the vehicle under test identified by the distributed optical fiber are acquired, the distributed optical fiber acquisition database is updated, and the second relative position and second detection time of the vehicle under test are generated.

[0009] Set a similarity threshold and determine the similarity between the first relative position and the second relative position. If the similarity threshold is exceeded, the vehicle to be tested identified by the camera and the distributed optical fiber is identified as the same vehicle, i.e., the target vehicle.

[0010] The pixel distance of the target vehicle is obtained based on the first relative position and the first detection time, and the movement distance of the target vehicle is obtained based on the second relative position and the second detection time. Within the detection area, the correspondence between pixel coordinates and world coordinates is constructed through the correspondence between pixel distance and movement distance, and the speed of the target vehicle is obtained.

[0011] As a further technical solution, the detection area is defined as follows: the starting line of the distributed optical fiber within the camera's acquisition area is set as the starting line of the detection area, and the ending line of the distributed optical fiber is set as the ending line of the detection area.

[0012] As a further technical solution, after acquiring the real-time data of the vehicle under test identified by the camera, a first number, a first target type, pixel coordinates, and a target detection box of the vehicle under test are also generated; after acquiring the real-time data of the vehicle under test identified by the distributed optical fiber, a second number and a second target type of the vehicle under test are also generated.

[0013] As a further technical solution, the method for identifying the vehicle under test detected by the camera and the distributed optical fiber as the same vehicle is as follows: creating a matrix, constructing two dynamic programming tables of the same size as the matrix, namely a first dynamic programming table and a second dynamic programming table; progressively updating the first dynamic programming table and the second dynamic programming table; calculating the similarity between the first relative position and the second relative position based on the first dynamic programming table and the second dynamic programming table, and if it is less than the similarity threshold, then identifying the vehicle under test detected by the camera and the distributed optical fiber as the same vehicle, i.e., the target vehicle.

[0014] As a further technical solution, the specific method for gradually updating the first dynamic programming table and the second dynamic programming table is as follows: starting from the starting point, gradually update the detection time data in the first dynamic programming table, update the first dynamic programming table based on the detection time data acquired by the camera in real time, update and calculate the distance data between the two targets when reaching each point by combining the detection time data in the first dynamic programming table, and update the second dynamic programming table.

[0015] As a further technical solution, the specific method for obtaining the movement distance of the target vehicle based on the second relative position and the second detection time is as follows:

[0016] Where L represents the movement distance, T1 and T2 represent different second detection times, and (X,Y) represents the second relative position.

[0017] As a further technical solution, the specific method for obtaining the pixel distance of the target vehicle based on the first relative position and the first detection time is as follows:

[0018] Where XS represents the pixel distance, t1 and t2 represent different first detection times, and (x', y') represents the pixel coordinates.

[0019] Secondly, the present invention provides a vehicle speed measurement system based on multi-device fusion, comprising: a database construction model configured to: acquire vehicle data to be measured from cameras and distributed optical fiber acquisition respectively, and construct a camera acquisition database and a distributed optical fiber acquisition database accordingly;

[0020] The data recognition model is configured to: set a detection area within the overlapping area of ​​the camera and the distributed optical fiber acquisition; within the detection area, acquire real-time data of the vehicle under test identified by the camera, update the camera acquisition database, and generate a first relative position and a first detection time of the vehicle under test; acquire real-time data of the vehicle under test identified by the distributed optical fiber, update the distributed optical fiber acquisition database, and generate a second relative position and a second detection time of the vehicle under test.

[0021] The vehicle fusion model is configured to: set a similarity threshold and determine the similarity between the first relative position and the second relative position. If the similarity threshold is exceeded, the vehicle identified by the camera and the distributed optical fiber is considered to be the same vehicle, i.e., the target vehicle.

[0022] The coordinate transformation and velocity measurement model is configured to: obtain the pixel distance of the target vehicle based on the first relative position and the first detection time; obtain the movement distance of the target vehicle based on the second relative position and the second detection time; and construct the correspondence between pixel coordinates and world coordinates within the detection area through the correspondence between pixel distance and movement distance, and obtain the velocity of the target vehicle.

[0023] One or more technical solutions of the present invention have the following beneficial effects:

[0024] (1) This invention combines the advantages of precise positioning using distributed fiber optic equipment and target detection using deep learning algorithms, solving the problem of inaccurate pixel-to-world coordinate mapping in speed measurement using traditional deep learning algorithms. Distributed fiber optic equipment can provide precise location data, assisting deep learning algorithms in more accurately calculating vehicle movement distance and pixel distance, thereby significantly improving the accuracy of vehicle speed and distance measurement and providing more reliable data for traffic management.

[0025] (2) This invention solves the technical problem that existing speed measurement technologies relying on single devices perform poorly in complex traffic scenarios. By integrating multiple devices (cameras and distributed optical fibers), this invention enables the vehicle speed measurement method to stably and comprehensively acquire vehicle data even in complex environments. Furthermore, through the comprehensive analysis and processing of multi-source data using deep learning algorithms, the accuracy of speed measurement is not significantly affected, thus greatly enhancing the system's adaptability in complex scenarios. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a flowchart of the method according to Embodiment 1 of the present invention;

[0028] Figure 2 This is a schematic diagram of coordinate system transformation according to Embodiment 1 of the present invention;

[0029] Figure 3 This is a structural diagram of the deep learning algorithm according to Embodiment 1 of the present invention;

[0030] Figure 4 This is a diagram of the ResNet network structure according to Embodiment 1 of the present invention. Detailed Implementation

[0031] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0032] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0033] Example 1

[0034] like Figure 1 As shown, the present invention provides a vehicle speed measurement method based on multi-device fusion, specifically including the following:

[0035] (I) Building the Database

[0036] By fixing the camera equipment directly above the road, all lanes can be included in the target area of ​​the camera equipment. The distributed fiber optic equipment is installed on a section of the road within the target area of ​​the camera equipment for auxiliary positioning. The data of the vehicles under test collected by the camera and the distributed fiber optic equipment are acquired respectively, and the camera acquisition database and the distributed fiber optic acquisition database are constructed accordingly.

[0037] The vehicle being tested, captured by the camera, is assigned a unique identification number in the camera's data database. This number serves as a data entry, and the fields in this data entry include:

[0038] 1. First ID: This field represents the identification number of the vehicle under test captured by the camera.

[0039] 2. First target type: This field indicates the detection and recognition of different vehicle models by existing deep learning algorithms.

[0040] 3. First relative position: This field represents the coordinates of the vehicle under test in the coordinate system.

[0041] After the camera equipment tracks the vehicle under test, it can provide the displacement vector of the geometric center point of the vehicle under test in a two-dimensional coordinate system within a defined area:

[0042] P CA (t)=x(t)i+y(t);

[0043] Where i and j are unit vectors along the x and y axes respectively, then at time t0, the first relative position of the vehicle under test is (x, y).

[0044] 4. Pixel coordinates: This field represents the image pixel coordinates (x', y') after transformation based on relative position coordinates.

[0045] 5. Target Detection Box: This field represents the detection box information for real-time tracking of the vehicle under test based on deep learning technology.

[0046] 6. First detection time: This field indicates the time when the vehicle under test was captured by the camera.

[0047] The vehicle under test, acquired through distributed fiber optic data collection, is assigned a unique identification number in the database as a data entry. This data entry includes the following fields:

[0048] 1. Second number: The identification number of the vehicle under test acquired by the distributed optical fiber.

[0049] 2. Second target type: This field represents the device type estimate for the vehicle under test obtained by the existing algorithm.

[0050] 3. Second relative position: This field indicates the coordinates of the vehicle under test in the coordinate system.

[0051] After the distributed optical fiber tracks the vehicle under test, it can provide the displacement vector of the target in a one-dimensional coordinate system:

[0052] P DAS (X) = X(t)i;

[0053] Where i is a unit vector along the X-axis. Then, at time t0, the relative position of the vehicle under test along the X-axis is X.

[0054] After uniformly laying distributed optical fibers in the test section, the y-direction of each fiber is numerically calibrated, such as... Figure 2 As shown, the displacement vector of the vehicle under test in the Y direction in a one-dimensional coordinate system can be given by assigning different weights:

[0055] P DAS (Y)=(αY1+βY2)(t)j;

[0056] Where j is a unit vector along the Y-axis, Y1 is fiber 1 through which the vehicle under test interacts, and α is the weight value of the corresponding fiber; Y2 is fiber 2 through which the vehicle under test interacts, and β is the weight value of the corresponding fiber; then at time t0, the relative position of the vehicle under test along the y-axis is Y. Then at time t0, the second relative position of the vehicle under test is (X,Y).

[0057] 4. Second detection time: This field indicates the time when the vehicle under test was detected by the distributed optical fiber.

[0058] (II) Coordinate System Matching

[0059] like Figure 2 As shown, because the camera equipment is installed directly above the road with a large detection range, while the distributed fiber optic equipment, as an auxiliary device, only needs to be installed on a section of the road within the target area of ​​the camera, there is an overlapping area between the distributed fiber optic monitoring equipment and the camera equipment. This area is crucial for transferring the tracking of the vehicle to be detected from the distributed fiber optic equipment to the camera equipment (or vice versa). The first step is to match the coordinate systems of the two devices in the overlapping area. Taking a camera equipment installed at an intersection and the fiber optic cable within its detection range as an example, a detection area is set within the overlapping area of ​​the camera and the distributed fiber optic cable. The detection area is set within the camera equipment, with the starting line of the distributed fiber optic cable within the camera's acquisition area set as the starting line of the detection area, and the ending line of the distributed fiber optic cable set as the ending line of the detection area. Specifically, as follows... Figure 2 As shown.

[0060] (III) Fusion of the vehicle under test

[0061] A similarity threshold is set, and the similarity between the first relative position and the second relative position is judged. If the similarity threshold is exceeded, the vehicle being tested identified by the camera and the distributed optical fiber is considered to be the same vehicle, i.e., the target vehicle. The specific steps are as follows:

[0062] S1: Create a matrix: First, based on the relative positions of each target and the detection time, calculate the distance and time between each target data point in the two targets, and construct a matrix.

[0063] S2: Construct Dynamic Programming Tables: Based on the matrix, construct two dynamic programming tables of the same size as the matrix, namely the first dynamic programming table and the second dynamic programming table; the first dynamic programming table records the detection time data from the starting point to the current point. The second dynamic programming table records the distance between the two targets at the current point.

[0064] S3: Update the dynamic programming table: Gradually update the first and second dynamic programming tables; starting from the starting point, gradually update the detection time data in the first dynamic programming table. Update the first dynamic programming table based on the detection time data acquired in real time by the camera. Combine the detection time data from the first dynamic programming table with the calculation of the distance between the two targets when they reach each point, and update the second dynamic programming table. The second dynamic programming table contains two relative position data: the relative position coordinates of the image and the relative position coordinates of the laser. The distance between the two targets is calculated using the Euclidean distance formula based on the two relative position coordinates.

[0065] The second dynamic programming table is updated by selecting the relative position data corresponding to the detection time point of the distributed optical fiber closest to the first dynamic programming table, and calculating the distance.

[0066] S4: Calculate the similarity between the first relative position and the second relative position based on the first dynamic programming table and the second dynamic programming table. If the similarity is less than the similarity threshold, then the vehicle detected by the camera and the distributed optical fiber is identified as the same vehicle, i.e., the target vehicle. The similarity is determined based on the similarity threshold. For example, if the similarity threshold is set to 5cm, and the distance between the two is less than 5cm, then they are identified as the same vehicle.

[0067] (iv) Establish the correspondence between pixel coordinates and world coordinates to complete the speed measurement of the target vehicle.

[0068] After fusion, if the target types of the vehicles under test are the same, the two sets of data are merged into one. The number and detection time are retained as the first number and the first detection time (the values ​​when the camera acquired the data); the relative positions are recorded as the first relative position and the second relative position respectively; at the same time, new fields of motion distance and pixel distance are added.

[0069] The movement distance is calculated using the Euclidean distance formula based on the second relative position and the second detection time. The movement distance is expressed as:

[0070] Where L represents the movement distance, T1 and T2 represent different second detection times, and (X,Y) represents the second relative position.

[0071] The pixel distance is calculated using the Euclidean distance formula based on the first relative position and the first detection time. The pixel distance is expressed as:

[0072] Where XS represents the pixel distance, t1 and t2 represent different first detection times, and (x', y') represents the pixel coordinates.

[0073] By detecting the correspondence between pixel distance and motion distance within the detection area, a deep learning algorithm is used to continuously calculate the mapping relationship between each pixel and the actual motion distance, construct the correspondence between pixel coordinates and world coordinates, achieve accurate conversion between pixel coordinates and world coordinates, and obtain the speed of the target vehicle, thus realizing the determination of vehicle speed based on multi-device fusion of deep learning algorithms.

[0074] The above mainly refers to images where the relative position coordinates of pixels of each target in the integrated image correspond one-to-one with the relative position coordinates calculated by distributed optical fibers. For example, if an image contains multiple targets, and the pixel coordinates and actual coordinates of each target are known, the pixel distance and actual distance between them can be calculated. Using this image as input data, deep learning is performed to calculate the relationship between pixel distance and motion distance in the image, as detailed below:

[0075] In this embodiment, as Figure 3 As shown, first, an image is acquired, which includes the pixel coordinates (x, y, y) of the target. ,y p ) and corresponding actual coordinates (x) r ,y r The image is then processed using a convolutional network to extract multimodal features, such as... Figure 4 As shown, a ResNet network is used, combined with a multi-scale feature pyramid structure to extract image features; a coordinate encoding network is used to normalize pixel coordinates while keeping the actual coordinates in their original units, and a coordinate network is constructed; then the image features and the coordinate network are fused, an attention mechanism is added, and the image features and coordinate features are concatenated using the channel dimension; finally, a distance field matrix (H×W) with the same resolution as the input image is formed, where each element value represents the actual motion distance (unit: meters) corresponding to the pixel position.

[0076] Example 2

[0077] This embodiment provides a vehicle speed measurement system based on multi-device fusion, which specifically includes the following modules:

[0078] The database construction model is configured to: acquire vehicle data from cameras and distributed optical fibers respectively, and construct corresponding camera acquisition databases and distributed optical fiber acquisition databases.

[0079] The data recognition model is configured to: set a detection area within the overlapping area of ​​the camera and the distributed optical fiber acquisition; within the detection area, acquire real-time data of the vehicle under test identified by the camera, update the camera acquisition database, and generate a first relative position and a first detection time of the vehicle under test; acquire real-time data of the vehicle under test identified by the distributed optical fiber, update the distributed optical fiber acquisition database, and generate a second relative position and a second detection time of the vehicle under test.

[0080] The vehicle fusion model is configured to: set a similarity threshold and determine the similarity between the first relative position and the second relative position. If the similarity threshold is exceeded, the vehicle identified by the camera and the distributed optical fiber is considered to be the same vehicle, i.e., the target vehicle.

[0081] The coordinate transformation and velocity measurement model is configured to: obtain the pixel distance of the target vehicle based on the first relative position and the first detection time; obtain the movement distance of the target vehicle based on the second relative position and the second detection time; and construct the correspondence between pixel coordinates and world coordinates within the detection area through the correspondence between pixel distance and movement distance, and obtain the velocity of the target vehicle.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A vehicle speed measurement method based on multi-device fusion, characterized in that, include: Data from the vehicle under test is acquired by cameras and distributed optical fiber, respectively, and corresponding camera acquisition database and distributed optical fiber acquisition database are constructed. A detection area is set within the overlapping area of ​​the camera and the distributed optical fiber acquisition. The detection area is defined as follows: the starting line of the distributed optical fiber within the camera acquisition area is set as the starting line of the detection area, and the ending line of the distributed optical fiber is set as the ending line of the detection area. Within the detection area, real-time data of the vehicle under test identified by the camera is acquired, the camera acquisition database is updated, and the first relative position and the first detection time of the vehicle under test are generated. The system acquires real-time data of the vehicle under test identified by the distributed optical fiber, updates the distributed optical fiber acquisition database, and generates the second relative position and second detection time of the vehicle under test. A similarity threshold is set, and the similarity between the first relative position and the second relative position is determined. If the similarity is less than the threshold, the vehicle detected by the camera and the distributed optical fiber is identified as the same vehicle. Specifically, a matrix is ​​created, and two dynamic programming tables of the same size as the matrix are constructed based on the matrix, namely the first dynamic programming table and the second dynamic programming table; the first dynamic programming table and the second dynamic programming table are updated step by step; the similarity between the first relative position and the second relative position is calculated based on the first dynamic programming table and the second dynamic programming table. If the similarity is less than the threshold, the vehicle detected by the camera and the distributed optical fiber is identified as the same vehicle, i.e., the target vehicle. The pixel distance of the target vehicle is obtained based on the first relative position and the first detection time, and the movement distance of the target vehicle is obtained based on the second relative position and the second detection time. Within the detection area, the correspondence between pixel coordinates and world coordinates is constructed through the correspondence between pixel distance and movement distance, and the speed of the target vehicle is obtained.

2. The vehicle speed measurement method based on multi-device fusion as described in claim 1, characterized in that, After acquiring the real-time data of the vehicle under test identified by the camera, a first number, a first target type, pixel coordinates, and a target detection box of the vehicle under test are also generated; after acquiring the real-time data of the vehicle under test identified by the distributed optical fiber, a second number and a second target type of the vehicle under test are also generated.

3. The vehicle speed measurement method based on multi-device fusion as described in claim 1, characterized in that, The specific method for gradually updating the first dynamic programming table and the second dynamic programming table is as follows: starting from the starting point, gradually update the detection time data in the first dynamic programming table, update the first dynamic programming table based on the detection time data obtained by the camera in real time, update and calculate the distance data between the two targets when they reach each point by combining the detection time data in the first dynamic programming table, and update the second dynamic programming table.

4. The vehicle speed measurement method based on multi-device fusion as described in claim 1, characterized in that, The specific method for obtaining the target vehicle's movement distance based on the second relative position and the second detection time is as follows: ;in, Indicates the distance traveled. and These represent different second detection times, (X) Y) represents the second relative position.

5. The vehicle speed measurement method based on multi-device fusion as described in claim 1, characterized in that, The specific method for obtaining the pixel distance of the target vehicle based on the first relative position and the first detection time is as follows: Where XS represents the pixel distance, and These represent different first detection times. Represents pixel coordinates.

6. A vehicle speed measurement system based on multi-device fusion, characterized in that, include: The database construction model is configured to: acquire vehicle data from cameras and distributed optical fibers respectively, and construct corresponding camera acquisition databases and distributed optical fiber acquisition databases. The data recognition model is configured to: set a detection area within the overlapping area of ​​the camera and the distributed optical fiber acquisition, wherein the detection area is defined as follows: the starting line of the distributed optical fiber within the camera acquisition area is set as the starting line of the detection area, and the ending line of the distributed optical fiber is set as the ending line of the detection area; within the detection area, real-time data of the vehicle under test identified by the camera is acquired, the camera acquisition database is updated, and the first relative position and first detection time of the vehicle under test are generated. The system acquires real-time data of the vehicle under test identified by the distributed optical fiber, updates the distributed optical fiber acquisition database, and generates the second relative position and second detection time of the vehicle under test. The vehicle fusion model is configured to: set a similarity threshold and determine the similarity between the first relative position and the second relative position. If the similarity is less than the similarity threshold, the vehicle identified by the camera and the distributed optical fiber is considered to be the same vehicle, i.e., the target vehicle. Specifically, the process involves: creating a matrix; constructing two dynamic programming tables of the same size as the matrix, namely a first dynamic programming table and a second dynamic programming table; progressively updating the first and second dynamic programming tables; calculating the similarity between the first and second relative positions based on the first and second dynamic programming tables; and determining that the similarity between the first and second relative positions is less than a similarity threshold, thus identifying the vehicle being tested by the camera and the distributed optical fiber as the same vehicle, i.e., the target vehicle. The coordinate transformation and velocity measurement model is configured to: obtain the pixel distance of the target vehicle based on the first relative position and the first detection time; obtain the movement distance of the target vehicle based on the second relative position and the second detection time; and construct the correspondence between pixel coordinates and world coordinates within the detection area through the correspondence between pixel distance and movement distance, and obtain the velocity of the target vehicle.

7. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the vehicle speed determination method based on multi-device fusion as described in any one of claims 1-5.

8. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the vehicle speed determination method based on multi-device fusion as described in any one of claims 1-5.

Images