Steel haulage vehicle identification system and method

Abstract

The application provides a steel carrying vehicle identification system and method, which comprises a scene acquisition module, a topological interconnection module, a vehicle identification module and a vehicle discrimination module; the scene acquisition module is used for setting a recycled steel automatic grading area, collecting vehicle pictures in the area and transmitting the pictures to the topological interconnection module; the topological interconnection module is used for connecting each scene acquisition module through a fiber switch and a fiber cable and transmitting the vehicle pictures collected by the scene acquisition module to a host computer; the vehicle identification module is used for calling the vehicle pictures stored in the host computer, extracting feature points in the vehicle pictures through a convolutional neural network and judging whether a carrying vehicle exists in the automatic grading area; and the vehicle discrimination module is used for extracting the vehicle pictures stored in the host computer when the carrying vehicle is identified, collecting and analyzing suspended license plate information through an image enhancement technology, obtaining a discrimination result and transmitting the result to the host computer. The application adopts the vehicle identification module to analyze the collected pictures, adopts the vehicle discrimination module to lock the carrying vehicle and realizes automatic identification of the carrying vehicle entering and leaving a warehouse.

Description

Technical Field

[0001] This invention relates to the field of vehicle identification technology, and more specifically, to a steel transport vehicle identification system and method. Background Technology

[0002] Currently, many vehicle recognition methods rely on cameras to capture images, setting up cameras at various entrances and exits to record vehicles entering and exiting designated areas. This method often suffers from errors in vehicle identification, such as misidentification or omission. For example, if a vehicle changes its license plate when entering or exiting the area, the information about its entry and exit will be incorrect. Furthermore, with the development of camera technology and the continuous expansion of camera coverage areas, existing vehicle recognition methods only record the time of vehicle entry and exit, resulting in a significant waste of camera coverage space.

[0003] Patent document CN107730897B (application number: CN201610824800.4) discloses a vehicle identification method, which includes the following steps: obtaining the license plate number of the front of the vehicle; controlling a camera device to record a video of the vehicle passing through a gate; identifying the license plate number of the rear of the vehicle based on the video; if the license plate number of the front of the vehicle is the same as the license plate number of the rear of the vehicle, then determining the license plate number of the front or rear of the vehicle as the vehicle's identity information. However, this patent only uses the vehicle license plate as the unique identity information of the vehicle, and cannot handle anomalies when the license plates of the front and rear of the vehicle are inconsistent.

[0004] Patent document CN105678275A (application number: CN201610027179.9) discloses a vehicle recognition method for automatically identifying vehicle types from vehicle images. The method involves inputting the PCA feature values ​​of vehicle images into a trained backpropagation (BP) neural network, which then outputs the recognition result. The training method for the BP neural network is as follows: first, a sample library containing several training sample images covering all vehicle types is established; then, the PCA feature values ​​of each training sample image are preprocessed and extracted; next, the PCA feature values ​​of each training sample image are input into the BP neural network, and the network is trained so that it outputs the correct recognition result after inputting the PCA feature values ​​of each training sample image. However, this patent can only identify regular vehicles and cannot identify or handle abnormal vehicles. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the purpose of this invention is to provide a steel transport vehicle identification system and method.

[0006] The steel transport vehicle identification system provided by the present invention includes:

[0007] Scene acquisition module: Sets up an automatic classification area for recycled steel, collects vehicle images within the area, and transmits them to the topology interconnection module;

[0008] Topology Interconnection Module: Connects various scene acquisition modules through fiber optic switches and fiber optic cables, and transmits the vehicle images captured by the scene acquisition modules to the host computer;

[0009] Vehicle recognition module: retrieves vehicle images stored in the host computer, extracts feature points from the vehicle images through a convolutional neural network, and determines whether there are transport vehicles in the automatic classification area;

[0010] Vehicle identification module: When a vehicle is identified, the module extracts the vehicle image stored in the host computer, collects and analyzes the license plate information through image enhancement technology, obtains the identification result, and transmits it to the host computer.

[0011] Preferred options also include:

[0012] Status detection module: Based on the 3D point cloud data and vehicle coordinates stored in the host computer, the module judges the operating status of the transport vehicles in the automatic classification area and uploads the status information of the transport vehicles to the host computer. The status includes: full load status, empty load status, moving status and stopped status.

[0013] Dynamic tracking module: Based on the vehicle coordinates fed back by the vehicle recognition module, it retrieves the identification results of the same transport vehicle in the scene from the vehicle discrimination module, and combines the status feedback from the status detection module to track and record the journey of the transport vehicle into the work area, and transmits the trajectory information to the host computer.

[0014] Preferred options also include:

[0015] Path drawing module: stitches together the automatically classified area images collected by the scene acquisition module to form a map, draws the map based on the vehicle trajectory information stored in the host computer, and transmits the stitched information containing the map and trajectory to the host computer.

[0016] Report generation module: Retrieves vehicle characteristics, status information, trajectory information and time information stored in the host computer to synthesize reports, and transmits the reports to the host computer for storage.

[0017] Preferably, the scene acquisition module includes:

[0018] The captured vehicle images are compressed. When there are no vehicles in the automatically classified area, the system enters a preset low bitrate camera mode, and when vehicles are detected in the area, the system enters a preset high bitrate camera mode.

[0019] Point cloud data of the scene within the automatically classified area is collected by LiDAR, the 3D point cloud data is mapped onto the 2D screen, and the mapped point cloud data is transmitted to the topology interconnection module.

[0020] Preferably, the image enhancement techniques include: Laplacian enhancement, Wiener filtering, and degradation estimation;

[0021] When the vehicle recognition module confirms the presence of a vehicle within the automatic classification area, it records the vehicle's feature points, including the vehicle's appearance, color, number of wheels, and license plate information. These features are then compared with the vehicle features identified during automatic classification to obtain the vehicle identification result, which is then fed back to the host computer.

[0022] The steel transport vehicle identification method provided by the present invention includes:

[0023] Scene acquisition steps: Set up an automatic classification area for recycled steel and collect vehicle images within the area;

[0024] Topology interconnection steps: Connect the acquisition device through fiber optic switch and fiber optic cable, and transmit the acquired vehicle images to the host computer;

[0025] Vehicle recognition steps: Retrieve vehicle images stored in the host computer, extract feature points from the vehicle images through a convolutional neural network, and determine whether there are any vehicles in the automatic classification area;

[0026] Vehicle identification steps: When a vehicle is identified, the vehicle image stored in the host computer is extracted, and the license plate information is collected and analyzed using image enhancement technology to obtain the identification result and transmit it to the host computer.

[0027] Preferred options also include:

[0028] Status detection steps: Based on the 3D point cloud data and vehicle coordinates stored in the host computer, the operating status of the transport vehicles in the automatic classification area is judged, and the status information of the transport vehicles is uploaded to the host computer. The status includes: full load status, empty load status, moving status and stopped status.

[0029] Dynamic tracking steps: Based on the vehicle coordinates obtained from vehicle identification, the judgment results obtained from vehicle discrimination, and the vehicle status obtained from status detection, the journey of the transport vehicle into the work area is tracked and recorded, and the trajectory information is transmitted to the host computer.

[0030] Preferred options also include:

[0031] Path drawing steps: stitch together the automatically classified area images of the scene to form a map, draw the map based on the vehicle trajectory information stored in the host computer, and transmit the stitched information containing the map and trajectory to the host computer.

[0032] Report generation steps: Retrieve vehicle characteristics, status information, trajectory information and time information stored in the host computer to synthesize a report, and then transmit the report to the host computer for storage.

[0033] Preferably, the scene acquisition step includes:

[0034] The captured vehicle images are compressed. When there are no vehicles in the automatically classified area, the system enters a preset low bitrate camera mode, and when vehicles are detected in the area, the system enters a preset high bitrate camera mode.

[0035] Point cloud data of the scene within the automatically classified area is collected by LiDAR, and the 3D point cloud data is mapped onto a 2D screen. The data is then compared with the collected vehicle images to verify the vehicle's position.

[0036] Preferably, the image enhancement techniques include: Laplacian enhancement, Wiener filtering, and degradation estimation;

[0037] When a vehicle is identified and confirmed to be in the automatic classification area, the vehicle's feature points are recorded, including its appearance, color, number of wheels, and license plate information. These features are then compared with the vehicle features identified during automatic classification to obtain the vehicle identification result, which is then fed back to the host computer.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] (1) The present invention uses a vehicle identification module to analyze the collected images and a vehicle discrimination module to lock the transport vehicle, thereby realizing the automatic identification of the transport vehicle entering and leaving the warehouse within the automatic classification area;

[0040] (2) By adopting a structure that combines lidar and camera, the present invention collects two-dimensional and three-dimensional information simultaneously, and collects the feature information of the transport vehicle entering the automatic classification area in a multi-dimensional way, thus solving the problems of vehicle misjudgment, omission, and wrong judgment.

[0041] (3) This invention summarizes the vehicle information collected by the host computer and generates vehicle entry and exit reports in a digital way, and records and summarizes the vehicle travel trajectory and unloading and other business-related information, thus solving the problem that the transportation route of the vehicle in the automatic classification area is untraceable.

[0042] (4) This invention connects the coordinates of the vehicle identification into a line, locks the transport vehicle in combination with the vehicle identification module, and records the movement trajectory of the vehicle in the automatic classification area, thus solving the problem of tracing the transport journey of the transport vehicle in the automatic classification area. Attached Figure Description

[0043] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0044] Figure 1 A flowchart for vehicle recognition;

[0045] Figure 2 This is a schematic diagram of point cloud data mapping. Detailed Implementation

[0046] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0047] Example:

[0048] The vehicle identification system for automatic grading of recycled steel provided by the present invention includes:

[0049] The scene acquisition module is used to acquire images of the automatically classified area for recycled steel, and to collect images of vehicles within the area. All acquired camera images are compressed using methods such as H.264, H.265, and MPEG. When no vehicles are detected within the area, the module enters a low-bitrate camera mode; when vehicles are detected, it enters a high-bitrate camera mode, transmitting images showing vehicles within the automatically classified area to the topology interconnection module. Point cloud data of the scene within the automatically classified area is collected using LiDAR. This 3D point cloud data is mapped onto a 2D image using a mapping formula, and the mapped point cloud data is then transmitted to the topology interconnection module.

[0050] The topology interconnection module connects each scene acquisition module through fiber optic switches and fiber optic cables, and transmits the images captured by the scene acquisition modules within the automatic classification area to the host computer.

[0051] The vehicle recognition module retrieves images of vehicles within the automatically classified area stored in the host computer, recorded by the scene acquisition module. This data is then processed by a convolutional neural network (CNN) to extract feature points from the images and determine the presence of a steel-carrying vehicle within the automatically classified area. Simultaneously, it retrieves 3D point cloud data of the area and combines it with the vehicle coordinates output by the neural network, converting this data into 3D point cloud data to verify the vehicle's presence. The vehicle recognition module then transmits the results of the automatically classified area's vehicle presence, along with the 2D and 3D coordinates of any present vehicles, to the host computer.

[0052] The vehicle identification module, when the vehicle recognition module reports to the host computer that a steel-carrying vehicle exists within the automatically classified area, analyzes the automatically classified area image stored in the host computer, extracts details of the front or rear of the vehicle, and collects the license plate information of the front or rear of the vehicle using image enhancement techniques. This collected license plate information is then transmitted to the host computer. Image enhancement techniques include Laplacian enhancement, Wiener filtering, and estimation degradation. Once the vehicle recognition module confirms the presence of a steel-carrying vehicle within the automatically classified area, it records the vehicle's feature points, including its appearance, color, number of wheels, and license plate information. By comparing these records, it ensures that the vehicle identified during automatic classification is the same vehicle, and then feeds the vehicle identification result back to the host computer.

[0053] The status detection module collects 3D point cloud data and vehicle coordinates stored in the host computer to determine the operating status of the transport vehicles within the automatic classification area. It analyzes the images of the automatic classification area stored in the host computer and uploads the transport vehicle status information fed back by the module to the host computer. The status includes: fully loaded, empty, moving, and stopped.

[0054] The dynamic tracking module uses the vehicle coordinates fed back by the vehicle recognition module to retrieve the identification results of the same transport vehicle in the scene from the vehicle discrimination module. Combined with the status feedback from the status detection module, it tracks and records the journey of the transport vehicle into the work area and transmits the trajectory information to the host computer.

[0055] The path drawing module stitches together the automatically classified area images collected by the scene acquisition module to form a map, collects and draws the vehicle trajectory information stored in the host computer, and transmits the stitched information containing the map and trajectory to the host computer.

[0056] The report generation module retrieves vehicle characteristics, status information, trajectory information, and time information stored in the host computer to synthesize reports, and then transmits the reports to the host computer for storage. This ensures that users can monitor the trajectory formation of each vehicle within the automatic classification area and the corresponding time status information.

[0057] like Figure 1 The scene acquisition module collects information on the operation within the automatically classified area. The camera captures two-dimensional images, while the LiDAR collects three-dimensional point cloud data within the automatically classified area. For example... Figure 2 The 2D images and 3D point cloud data acquired by the scene acquisition module need to be mapped. The mapping formula is as follows:

[0058]

[0059] (u,v) are the pixel coordinates in the 2D image, (X,Y,Z) are the coordinates of the corresponding pixel in the 3D point cloud, and (f u ,fv (u0, v0) is the camera intrinsic parameter matrix, and (R, t) is the rigid body transformation matrix that maps the origin of the 3D point cloud to the coordinate system corresponding to the camera. After transformation and correction, the 3D point cloud data is mapped onto the 2D image.

[0060] like Figure 1The scene acquisition module collects 2D images and 3D point cloud data within the automatically classified area and transmits the data to the topology interconnection module. To ensure real-time and synchronous data processing, the 2D images and 3D point clouds are transmitted to the host computer via low-latency, high-response fiber optic cables. The host computer stores the real-time 2D images and 3D point cloud data collected by the scene acquisition module. The vehicle recognition module retrieves the real-time 2D images and 3D point cloud data stored in the host computer. The 2D images are first preprocessed by the input layer using mean reduction, normalization, and SVD dimensionality reduction. After processing, the images are processed by a convolutional neural network for recognition. The convolutional layer uses the receptive field to locally perceive details within the image, and then the pooling layer downsamples the image to reduce overfitting. The 3D point cloud data retrieves the vehicle coordinates output by the neural network and converts them into 3D point cloud data to verify the presence of a vehicle. When the vehicle recognition module determines that a vehicle exists within the automatic classification area, the host computer activates the vehicle discrimination module. This module retrieves real-time scene information stored on the host computer, zooms in on the scene acquisition module, and focuses the image on the front or rear of the vehicle facing the scene acquisition module, capturing the license plate information at either end. Simultaneously, by combining vehicle feature information and license plate information stored on the host computer, a multi-dimensional feature analysis confirms that the transport vehicle within the automatic classification area is the same vehicle, ensuring accurate vehicle identification. If the vehicle license plate does not match the vehicle features recorded by the vehicle discrimination module, an audible and visual alarm is triggered within the automatic classification area. The dynamic detection module combines the vehicle coordinates from the vehicle recognition module and the vehicle identity information from the discrimination module within the automatic classification area, feeding back the vehicle's trajectory to the host computer. After the vehicle recognition module determines that the transport vehicle has arrived at the automatic classification point and the vehicle identification module extracts the vehicle's feature information, the status detection module is activated. It retrieves the 3D point cloud data stored in the host computer and, combined with the 2D vehicle coordinates fed back by the vehicle recognition module, performs an inverse transformation using a mapping formula to convert the vehicle coordinates into 3D point cloud coordinates. This coordinate system is then used to search for the loading status inside the vehicle's cargo bed and feeds the loading status back to the host computer. Simultaneously, the 2D image is compared with the vehicle coordinates and features fed back by the vehicle recognition module in the host computer, and the status detection module determines the transport vehicle's movement status. The status detection module transmits the transport vehicle's status information to the host computer. When the vehicle recognition module reports the coordinates of the transport vehicle that entered the automatic classification area and then left the automatic classification area, the path drawing module is activated. It records the transport vehicle's movement trajectory within the automatic classification area, combining vehicle identity information and time information, and transmits the recorded information to the host computer. The report generation module combines the vehicle's feature information, status information, and time information into a single report, allowing users to easily review the transport vehicle's movement information within the automatic classification area.

[0061] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0062] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A steel hauler vehicle identification system, characterized by, include: Scene acquisition module: Sets up an automatic classification area for recycled steel, collects vehicle images within the area, and transmits them to the topology interconnection module; Topology Interconnection Module: Connects various scene acquisition modules through fiber optic switches and fiber optic cables, and transmits the vehicle images captured by the scene acquisition modules to the host computer; Vehicle recognition module: retrieves vehicle images stored in the host computer, extracts feature points from the vehicle images through a convolutional neural network, and determines whether there are transport vehicles in the automatic classification area; Vehicle identification module: When a vehicle is identified, the module extracts the vehicle image stored in the host computer, collects and analyzes the license plate information through image enhancement technology, obtains the identification result, and transmits it to the host computer. Status detection module: Based on the 3D point cloud data and vehicle coordinates stored in the host computer, the module judges the operating status of the transport vehicles in the automatic classification area and uploads the status information of the transport vehicles to the host computer. The status includes: full load status, empty load status, moving status and stopped status. Dynamic tracking module: Based on the vehicle coordinates fed back by the vehicle recognition module, it retrieves the identification results of the same transport vehicle in the scene from the vehicle discrimination module, and combines the status fed back by the status detection module to track and record the journey of the transport vehicle into the work area, and transmits the trajectory information to the host computer. Path drawing module: stitches together the automatically classified area images collected by the scene acquisition module to form a map, draws the map based on the vehicle trajectory information stored in the host computer, and transmits the stitched information containing the map and trajectory to the host computer. Report generation module: retrieves vehicle characteristics, status information, trajectory information and time information stored in the host computer to synthesize reports, and transmits the reports to the host computer for storage; The captured vehicle images are compressed. When there are no vehicles in the automatically classified area, the system enters a preset low bitrate camera mode, and when vehicles are detected in the area, the system enters a preset high bitrate camera mode. Point cloud data of the scene within the automatic classification area is collected by LiDAR, the 3D point cloud data is mapped onto the 2D screen, and the mapped point cloud data is transmitted to the topology interconnection module. The image enhancement techniques include: Laplacian enhancement, Wiener filtering, and degradation estimation. When the vehicle recognition module confirms the presence of a vehicle within the automatic classification area, it records the vehicle's feature points, including the vehicle's appearance, color, number of wheels, and license plate information. These features are then compared with the vehicle features identified during automatic classification to obtain the vehicle identification result, which is then fed back to the host computer.

2. A steel transport vehicle identification method characterized by, include: Scene acquisition steps: Set up an automatic classification area for recycled steel and collect vehicle images within the area; Topology interconnection steps: Connect the acquisition device through fiber optic switch and fiber optic cable, and transmit the acquired vehicle images to the host computer; Vehicle recognition steps: Retrieve vehicle images stored in the host computer, extract feature points from the vehicle images through a convolutional neural network, and determine whether there are any vehicles in the automatic classification area; Vehicle identification steps: When a transport vehicle is identified, the vehicle image stored in the host computer is extracted, and the license plate information is collected and analyzed through image enhancement technology to obtain the identification result and transmit it to the host computer. Also includes: The state detection step: according to the three-dimensional point cloud data stored in the host computer and the vehicle coordinates, the working state of the carrying vehicle in the automatic grading area is judged, and the vehicle state information is uploaded to the host computer. The state includes: full load state, empty load state, motion state and stop state; The dynamic tracking step: according to the vehicle coordinates obtained by vehicle recognition, the discrimination result obtained by vehicle discrimination and the vehicle state obtained by state detection, the journey of the carrying vehicle entering the working area is tracked and recorded, and the trajectory information is transmitted to the host computer; Also includes: The path drawing step: the image of the automatic grading area is spliced to form a map, and the trajectory information of the carrying vehicle stored in the host computer is drawn, and the splicing information containing the map and the trajectory is transmitted to the host computer; The report generation step: the vehicle features, state information, trajectory information and time information stored in the host computer are called to synthesize a report, and the report is transmitted to the host computer for storage; The scene acquisition step includes: The collected vehicle picture is compressed, and when there is no vehicle in the automatic grading area, the preset low code rate camera mode is entered, and when a vehicle is detected in the area, the preset high code rate camera mode is entered; The point cloud data of the scene in the automatic grading area is collected by laser radar, the three-dimensional point cloud data is mapped to a two-dimensional picture, and the vehicle position is verified by comparing with the collected vehicle picture; The image enhancement technology includes: laplace enhancement, wiener filtering and estimation degradation; When the vehicle recognition confirms that there is a carrying vehicle in the automatic grading area, the vehicle feature points are recorded, including vehicle appearance, color, wheel number and license plate information, which are compared with the vehicle features in the automatic grading to obtain the vehicle discrimination result and feedback to the host computer.

Images