Park vehicle real-time tracking system

Abstract

According to the park vehicle real-time tracking system provided by the invention, the vehicles in the park are monitored in real time through the unmanned aerial vehicle constellation, the unmanned aerial vehicles are deployed in the air of the park, the problem of insufficient point locations does not exist, a small number of unmanned aerial vehicles can cover a vast monitoring area, the deployment cost is low, the problem of mutual shielding between vehicles and between vehicles and objects is avoided, and the real-time tracking of the vehicles in the park is realized. According to the method, the park ground can be continuously monitored in the two dimensions of time and space, the problem of mutual shielding between vehicles and between vehicles and objects is effectively solved, the processing equipment accurately determines the real-time position of each vehicle based on pictures transmitted back in real time by the unmanned aerial vehicle, driving track data of each vehicle is formed, and the real-time position of each vehicle is obtained. The latest position of each vehicle can be obtained by inquiring the database, convenience and high efficiency are achieved, processing equipment can effectively distinguish different vehicles in pictures based on the unique advantage of the shooting angle, and meanwhile the system does not depend on cooperation of drivers.

Description

Technical Field

[0001] The invention belongs to the technical field of vehicle tracking, and in particular relates to a real-time tracking system for park vehicles. Background Art

[0002] There are many vehicles entering and leaving large industrial parks every day. Some application fields need to be able to grasp the precise location of vehicles in real time. However, the current technology for managing vehicles in the park is still relatively extensive. After the vehicle enters the park, there is a lack of an effective technical system to track the driving trajectory of each vehicle in real time, and it is impossible to achieve advanced functions such as behavior monitoring and risk warning. Taking the port scene as an example, there are often tens of thousands of container trucks in large port areas at the same time. When the park management needs to find a specific truck, it can only obtain the time and location of the vehicle last captured by the system by checking the video surveillance. The search process is time-consuming and laborious, and the information obtained is delayed and does not represent the latest location of the vehicle.

[0003] Some existing technologies use wireless technologies such as ETC / RFID to collect the location of vehicles, hoping to at least solve the problem of inconvenient video surveillance access to a certain extent. Such technologies require issuing cards to vehicles when they enter the park and collecting them when they leave the park. The inconveniences are that, first, the cost of issuing cards is introduced, and once the cards are lost, disputes are likely to arise. Second, the reliability of such technologies is not high, and they often miss detections due to unstable signals. Third, they rely on the active cooperation of the driver. If the driver does not cooperate, this technology is likely to fail.

[0004] Some existing technologies rely on the existing video surveillance network in the park to capture the license plate of the vehicle, and try to grasp the real-time location of the vehicle through license plate recognition technology. This type of technology also has several defects. First, there are many technical limitations in video surveillance to shoot license plates. When the shooting angle is not suitable, the vehicles are blocked by each other, or the shooting distance is too far, it is impossible to shoot a license plate of qualified quality. Second, the conditions for deploying monitoring equipment in the park are limited, and it is impossible to provide sufficient points to deploy enough cameras.

[0005] Therefore, for large-scale closed industrial parks, achieving and maintaining real-time tracking of any vehicle's driving trajectory within a larger time and space range remains an unresolved issue. Summary of the invention

[0006] Based on this, in order to solve the above technical problems, a real-time tracking system for park vehicles is provided.

[0007] The technical solution adopted by the present invention is as follows: A real-time tracking system for park vehicles, characterized by comprising: A plurality of drones hovering in a fixed position over the park, the plurality of drones forming a drone constellation covering the park, the drones being provided with bird's-eye view cameras, and taking a bird's-eye view picture of the park in real time through the bird's-eye view cameras and sending the picture to a processing device; An access control camera is installed at the entrance and exit of the park, and the access control camera takes a frontal photo of a vehicle at the access control point and sends it to the processing device; The processing device extracts vehicle features from the front photo and creates a file for the corresponding vehicle in a database, wherein the file includes vehicle features and driving trajectory data of the corresponding vehicle, wherein the driving trajectory data is a collection of vehicle positions, and an initial position thereof is the longitude and latitude coordinates of the corresponding entrance and exit; the processing device determines the longitude and latitude coordinates of the vehicle target in the bird's-eye view picture, finds the corresponding vehicle in the database according to the longitude and latitude coordinates and the vehicle features, and adds the longitude and latitude coordinates to the driving trajectory data of the corresponding vehicle in the database.

[0008] The present invention monitors the vehicles in the park in real time through a drone constellation. Since the drones are deployed in the sky above the park, there is no problem of insufficient points. A small number of drones can cover a large monitoring area, and the deployment cost is not high. The problem of mutual obstruction between vehicles and vehicles, and vehicles and objects is avoided. Uninterrupted monitoring of the park ground can be achieved in both time and space dimensions, and the problem of mutual obstruction between vehicles and vehicles and objects is effectively solved. The processing equipment accurately determines the real-time position of each vehicle based on the pictures sent back by the drone in real time, and forms the driving trajectory data of each vehicle. The latest position of each vehicle can be obtained by querying the database. It is convenient and efficient. Based on the unique advantage of shooting angle, the processing equipment can effectively distinguish different vehicles in the picture. At the same time, the system does not rely on the cooperation of the driver. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments: Figure 1 A schematic diagram of the structure of a real-time tracking system for park vehicles provided by an embodiment of the present invention; Figure 2 Schematic diagram of the point deployment of the drone constellation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The following will illustrate the implementation methods of the present invention in conjunction with the drawings in the specification. It should be noted that the implementation methods involved in this specification are not exhaustive and do not represent the only implementation methods of the present invention. The following corresponding embodiments are only for the purpose of clearly illustrating the invention content of the patent of the present invention, and are not intended to limit its implementation methods. For ordinary technicians in this field, different forms of changes and modifications can be made on the basis of the description of this embodiment. All obvious changes or modifications that belong to the technical concept and invention content of the present invention are also within the scope of protection of the present invention.

[0011] like Figure 1 As shown, an embodiment of the present application provides a real-time tracking system for campus vehicles, including multiple drones 110, access control cameras 120, a processing device 130, a display device 140 and a drone control device 150.

[0012] Among them, the drone 110, the access control camera 120 and the processing device 130 are connected in communication, and the drone 110 is connected in communication with the drone control device 150. There are two commonly used communication connection technologies. One is that the communicating parties use the finished 4G / 5G communication module to transmit data through the Ethernet protocol; the other is to use a dedicated industrial-grade wireless communication solution. The radio frequency is different from the mobile phone frequency. Generally, a private transmission protocol is used, which has a small communication delay and high fault tolerance. It is the preferred solution for ground-to-air communication.

[0013] Multiple drones 110 are hovering in a low altitude area above the park 2, and each drone 110 is on duty at a fixed point 21, see Figure 2 , forming a drone constellation covering Park 2. Since the drone constellation is deployed above Park 2, there is no problem of insufficient points.

[0014] The drone 110 is equipped with a bird's-eye view camera, which takes a bird's-eye view picture of the park 2 in real time and sends it to the processing device 130. The bird's-eye view camera has a high shooting position and a large camera coverage area. A small number of drones can cover a large monitoring area, avoiding the problem of mutual occlusion between vehicles and between vehicles and objects. The park ground can be continuously monitored in both time and space dimensions, effectively solving the problem of mutual occlusion between vehicles and between vehicles and objects. In addition, based on the unique advantage of the shooting angle, the processing device 130 can effectively distinguish different vehicles in the picture.

[0015] Among them, the drone 110 is a multi-rotor drone.

[0016] The access control camera 120 is arranged at the entrance and exit of the park 2 . When a vehicle stops and waits at the access control area, the access control camera 120 takes a frontal photo of the vehicle at the access control area and sends it to the processing device 130 .

[0017] The processing device 130 extracts vehicle features from the front photo through an artificial intelligence model and creates a file for the corresponding vehicle in the database. The file includes the vehicle features and driving trajectory data of the corresponding vehicle. The driving trajectory data is a collection of vehicle positions, and its initial position is the longitude and latitude coordinates of the corresponding entrance and exit. For example, if the vehicle enters Park 2 from the left entrance and exit, its initial position is the longitude and latitude coordinates of the left entrance and exit.

[0018] Among them, vehicle characteristics include license plate, model, logo, color and structure (the same below).

[0019] The artificial intelligence model can adopt the yolo series model, including yolo v5, yolo v8, yolo v10 and other versions.

[0020] The processing device 130 determines the longitude and latitude coordinates of the vehicle target in the bird's-eye view image, finds the corresponding vehicle in the database based on the longitude and latitude coordinates and vehicle features, and adds the longitude and latitude coordinates to the driving trajectory data of the corresponding vehicle in the database. It can be seen that the driving trajectory data is actually a time series of longitude and latitude coordinates.

[0021] The specific process of determining the latitude and longitude coordinates of the vehicle target in the bird's-eye view picture is as follows: S31a, pre-processing the bird's-eye view image to transform its viewing angle from the drone shooting angle to the vehicle target's upper viewing angle and align it with the reference image of the corresponding point.

[0022] Preprocessing includes two types of operations: image quality enhancement and geometric transformation of the original bird's-eye view images.

[0023] Image quality enhancement operations mainly include: brightness and color balance, noise reduction, sharpening, dehazing, anti-shake, contrast enhancement, etc.

[0024] The geometric transformation operations mainly include: removing lens distortion, perspective transformation, etc. The meaning of geometric transformation is to transform the image perspective from the drone shooting position to the top of the vehicle target, and transform the pixel density of the target in the image (the value depends on the physical distance from the target to the drone) to a pre-defined standard pixel density (that is, the pixel density that should be achieved when the drone is shooting at a specific distance above the target).

[0025] Therefore, a reference image is required in the preprocessing process. The reference image is taken by using a standard drone, at a standard spatial position (spatial coordinates and attitude pointing), and under a standard technical state to shoot the park. An image with the best image quality is selected from a large number of samples and determined as the reference image. When the drone 110 shoots real-life images under actual working conditions, the real-life images will also deviate from the reference images to varying degrees because the drone will deviate from the standard drone in terms of spatial position and technical parameters. After preprocessing, the real-life images taken by the drone can be corrected to be as close to the reference images as possible.

[0026] It should be pointed out that the reference pictures of different points need to be determined separately. For example, the reference picture of point A needs to be photographed and determined by a standard drone at the standard spatial position of point A. When the bird's-eye view picture taken at point A needs to be preprocessed, the reference picture of point A needs to be used.

[0027] S31b, obtaining the pixel coordinates of the vehicle target from the preprocessed image. The pixel coordinates of the vehicle target in the image can be obtained through the target recognition model.

[0028] S31c, determine the latitude and longitude coordinates of the vehicle target by the following formula: G=Inv(M)*P, where G represents the latitude and longitude coordinates of the vehicle target, M represents the calibrated transformation matrix (3*3), P represents the pixel coordinates of the vehicle target, and Inv() represents the inverse matrix of the matrix.

[0029] The obtained latitude and longitude coordinates can be accurate to centimeters.

[0030] It should also be pointed out that the transformation matrices of different points need to be determined separately. Taking point A as an example, select a standard drone and make it shoot the marker on the ground at the standard position of point A. The latitude and longitude coordinates of the marker can be obtained using a laser rangefinder, and the pixel coordinates of the marker can be measured using software. Therefore, the transformation matrix of point A can be determined based on the latitude and longitude coordinates and pixel coordinates of the marker. When it is necessary to determine the latitude and longitude coordinates of the vehicle target in the picture taken at point A, the transformation matrix of point A is used.

[0031] The specific process of finding the corresponding vehicle in the database based on the latitude and longitude coordinates and vehicle features is as follows: S32a, finding vehicles with similar positions from the database according to the latitude and longitude coordinates of the vehicle target to form a candidate vehicle set.

[0032] For example, a circle may be drawn with the latitude and longitude coordinates of the vehicle target as the center and a preset radius, and a vehicle whose latitude and longitude coordinates are within the circle may be found from the database as an approaching vehicle.

[0033] S32b. Extract vehicle features of the vehicle target through an artificial intelligence model.

[0034] S32c, calculating the similarity confidence between the vehicle characteristics of the vehicle target and the vehicle characteristics of each candidate vehicle in the candidate vehicle set.

[0035] Specifically, the vehicle features of the vehicle target and the vehicle features of the candidate vehicle can be represented as an N-dimensional vector (N=256 or 512 is recommended), the dot product of the two vectors is calculated and normalized, and the normalized dot product value is used as the similarity confidence.

[0036] S32d. From the candidate vehicles whose similarity confidence is greater than a confidence threshold (such as 0.9), select the candidate vehicle with the largest similarity confidence as the corresponding vehicle.

[0037] The display device 140 is signal-connected to the processing device 130, and an interactive interface can be provided on the display screen of the display device 140. The operator can select any vehicle on the display screen. The display device 140 responds to the operation, queries the driving trajectory data of the vehicle through the database, and draws the driving trajectory on the display screen.

[0038] The drone control device 150 is set in the drone center, which monitors the power of each drone in the drone constellation. When the power of a drone is lower than a threshold, it controls the drone to return to the drone center for charging, and controls the backup drones in the drone center to move to the corresponding location, thereby realizing 24-hour uninterrupted service for the drone constellation.

[0039] As can be seen from the above, a real-time tracking system for park vehicles provided in an embodiment of the present application monitors the vehicles in the park in real time through a drone constellation. Since the drones are deployed in the sky above the park, there is no problem of insufficient points. A small number of drones can cover a large monitoring area, and the deployment cost is not high. The problem of mutual obstruction between vehicles and vehicles, and vehicles and objects is avoided. Uninterrupted monitoring of the park ground can be achieved in both time and space dimensions, and the problem of mutual obstruction between vehicles and vehicles, and vehicles and objects can be effectively solved. The processing equipment accurately determines the real-time position of each vehicle based on the pictures sent back by the drone in real time, and forms the driving trajectory data of each vehicle. The latest position of each vehicle can be obtained by querying the database. It is convenient and efficient. Based on the unique advantage of the shooting angle, the processing equipment can effectively distinguish different vehicles in the picture. At the same time, the system does not rely on the cooperation of the driver.

[0040] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A real-time tracking system for park vehicles, characterized in that: include: A plurality of drones hovering in a fixed position over the park, the plurality of drones forming a drone constellation covering the park, the drones being provided with bird's-eye view cameras, and taking a bird's-eye view picture of the park in real time through the bird's-eye view cameras and sending the picture to a processing device; An access control camera is installed at the entrance and exit of the park, and the access control camera takes a frontal photo of a vehicle at the access control point and sends it to the processing device; The processing device extracts vehicle features from the front photo and creates a file for the corresponding vehicle in a database, wherein the file includes vehicle features and driving trajectory data of the corresponding vehicle, wherein the driving trajectory data is a collection of vehicle positions, and an initial position thereof is the longitude and latitude coordinates of the corresponding entrance and exit; the processing device determines the longitude and latitude coordinates of the vehicle target in the bird's-eye view picture, finds the corresponding vehicle in the database according to the longitude and latitude coordinates and the vehicle features, and adds the longitude and latitude coordinates to the driving trajectory data of the corresponding vehicle in the database.

2. A real-time tracking system for park vehicles according to claim 1, characterized in that: The UAV is a multi-rotor UAV.

3. A real-time tracking system for park vehicles according to claim 1, characterized in that: The processing device extracts vehicle features from the frontal photo, further comprising: The processing device extracts vehicle features from the front photo through an artificial intelligence model.

4. A real-time tracking system for park vehicles according to claim 1, characterized in that: The processing device determines the latitude and longitude coordinates of the vehicle target in the bird's-eye view picture, further comprising: Preprocessing the bird's-eye view image so that its viewing angle is transformed from the perspective of drone shooting to the perspective directly above the vehicle target and aligned with the reference image of the corresponding point; Get the pixel coordinates of the vehicle target from the preprocessed image; The latitude and longitude coordinates of the vehicle target are determined by the following formula: G=Inv(M)*P, where G represents the latitude and longitude coordinates of the vehicle target, M represents the calibrated transformation matrix, and P represents the pixel coordinates of the vehicle target.

5. A real-time tracking system for park vehicles according to claim 1, characterized in that: The method of finding a corresponding vehicle in the database according to the latitude and longitude coordinates and the vehicle features further includes: Find vehicles with similar positions from the database according to the latitude and longitude coordinates of the vehicle target to form a candidate vehicle set; Extracting vehicle features of vehicle targets; Calculating the similarity confidence between the vehicle feature of the vehicle target and the vehicle feature of each candidate vehicle in the candidate vehicle set; From the candidate vehicles whose similarity confidence is greater than the confidence threshold, the candidate vehicle with the greatest similarity confidence is selected as the corresponding vehicle.

6. A real-time tracking system for park vehicles according to claim 5, characterized in that: The calculating of the similarity confidence between the vehicle feature of the vehicle target and the vehicle feature of each candidate vehicle in the candidate vehicle set further comprises: The vehicle characteristics of the vehicle target and the vehicle characteristics of the candidate vehicle are respectively represented as an N-dimensional vector; The dot product of the two vectors is calculated and normalized, and the normalized dot product value is used as the similarity confidence.

7. A real-time tracking system for park vehicles according to any one of claims 1 to 6, characterized in that: The vehicle characteristics include license plate, model, logo, color and structure.

8. A real-time tracking system for park vehicles according to claim 7, characterized in that: A display device is also included, and the display device is signal-connected to the processing device.

9. A real-time tracking system for park vehicles according to claim 8, characterized in that: It also includes a drone control device, which monitors the power of each drone in the drone constellation. When the power of a drone is lower than a threshold, the drone is controlled to return to the drone center for charging, and the backup drones in the drone center are controlled to move to corresponding locations.

Images