A vehicle barrier automatic license plate recognition control system

Abstract

The present invention relates to a vehicle barrier automatic license plate recognition control system, which relates to the field of image data processing technology and comprises: a vehicle sensing module, which is used to sense whether there is a vehicle in a lane; a vehicle recognition module, which is used to collect vehicle infrared sensing information when sensing the presence of a vehicle in the lane, and predict the three-dimensional coordinates of the license plate of the vehicle to be recognized based on the vehicle infrared sensing information; an image acquisition module, which is used to judge whether to adjust the image acquisition posture based on the predicted three-dimensional coordinates of the license plate of the vehicle to be recognized; if so, adjust the posture of the image acquisition device based on the three-dimensional coordinates of the license plate of the vehicle to be recognized, and then collect the image of the vehicle to be recognized; if not, collect the image of the vehicle to be recognized; and a license plate recognition module, which is used to recognize the license plate information of the vehicle to be recognized based on the image of the vehicle to be recognized. The invention has the advantages of improving the intelligence level of automatic license plate recognition of the vehicle barrier, improving the accuracy of license plate recognition and the traffic efficiency of vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a vehicle barrier automatic license plate recognition control system. Background Art

[0002] License plate recognition (LPR) systems are an application of computer video image recognition technology to vehicle license plate identification. LPR is widely used in highway vehicle management and is the primary means of identifying vehicles in electronic toll collection systems, in conjunction with DSRC technology. LPR technology, combined with electronic toll collection systems, allows vehicles to pass through intersections without stopping, enabling automatic vehicle identification and toll collection. In parking lot management, to improve access efficiency, LPR is being used to establish unmanned express lanes for vehicles exempt from parking fees (such as trucks and internal free-pass vehicles). This allows for card-free, no-stop access and is transforming parking lot management. A barrier gate integrated with a LPR system, called an all-in-one, is a specialized device for managing access to restricted lanes on roads. It is primarily used in highway toll booths and parking lot systems to manage vehicle access and entry. Its small footprint and relatively low cost make it widely used.

[0003] After the license plate recognition system recognizes the license plate, it controls the barrier to open and let the motor vehicle in. In this process, license plate recognition technology is very important. Its function is to use image recognition technology to analyze the captured image through software to find the license plate number, and extract it as the vehicle identification code, so as to carry out management activities such as capturing violations and parking fees. However, due to the variety of motor vehicles, the hanging position of the license plate is also different (such as height), and the driving information of each vehicle at the entrance and exit of the parking lot is also different (such as the position of driving off the main road). The position of the camera at the entrance and exit of the parking lot is generally fixed and not easy to adjust. Therefore, during the photo shooting process, for different vehicles, due to angle problems, the license plate number is often not well identified, which will affect the efficiency of vehicle entry and exit.

[0004] Therefore, it is necessary to provide a vehicle barrier automatic license plate recognition control system to improve the intelligence level of vehicle barrier automatic license plate recognition, improve the accuracy of license plate recognition and vehicle traffic efficiency. Summary of the Invention

[0005] The present invention provides a vehicle barrier automatic license plate recognition control system, comprising: a vehicle sensing module for sensing whether there is a vehicle in a lane; a vehicle recognition module, comprising an infrared sensing unit and a parameter determination unit, wherein the infrared sensing unit comprises a plurality of X-axis infrared sensing devices arranged along the width direction of the barrier and a plurality of Y-axis infrared sensing devices arranged along the length direction of the barrier, the X-axis infrared sensing device comprises a plurality of X-axis infrared sensors arranged along the vertical direction, the Y-axis infrared sensing device comprises a plurality of Y-axis infrared sensors arranged along the vertical direction, and the infrared sensing unit is used to collect vehicle infrared data when the vehicle sensing module senses that there is a vehicle in the lane. Sensing information, the parameter determination unit is used to predict the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle infrared sensing information; the image acquisition module is used to determine whether to adjust the image acquisition posture based on the predicted three-dimensional coordinates of the license plate of the vehicle to be identified. If it is determined to adjust the image acquisition posture, a target image acquisition posture is generated based on the three-dimensional coordinates of the license plate of the vehicle to be identified, and the posture of the image acquisition device is adjusted based on the target image acquisition posture. After the posture is adjusted, the image of the vehicle to be identified is acquired. If it is determined not to adjust the image acquisition posture, the image of the vehicle to be identified is acquired; the license plate recognition module is used to identify the license plate information of the vehicle to be identified based on the image of the vehicle to be identified.

[0006] Furthermore, when the vehicle sensing module senses that there is a vehicle in the lane, the infrared sensing unit collects vehicle infrared sensing information, including: starting each of the Y-axis infrared sensors to collect the Y-axis obstacle distance; starting each of the X-axis infrared sensors to collect the X-axis obstacle distance, wherein the vehicle infrared sensing information includes at least the Y-axis obstacle distance and the X-axis obstacle distance.

[0007] Furthermore, the parameter determination unit predicts the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle infrared sensing information, including: determining the vehicle posture change parameters of the vehicle to be identified at the multiple continuous time points based on the Y-axis obstacle distance collected by each of the Y-axis infrared sensors at multiple continuous time points and the X-axis obstacle distance collected by each of the X-axis infrared sensors at multiple continuous time points; determining the target time point from the multiple continuous time points based on the vehicle posture change parameters of the vehicle to be identified at the multiple continuous time points; determining the vehicle model parameters of the vehicle to be identified based on the Y-axis obstacle distance collected by each of the Y-axis infrared sensors at the target time point and the X-axis obstacle distance collected by each of the X-axis infrared sensors at the target time point, wherein the vehicle model parameters of the vehicle to be identified include at least the maximum width, maximum length, maximum height and chassis height of the vehicle; and predicting the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle model parameters of the vehicle to be identified.

[0008] Furthermore, the parameter determination unit determines the vehicle posture change parameters of the vehicle to be identified at the multiple consecutive time points based on the Y-axis obstacle distances collected by each of the Y-axis infrared sensors at the multiple consecutive time points and the X-axis obstacle distances collected by each of the X-axis infrared sensors at the multiple consecutive time points, including: for each of the time points, based on the Y-axis obstacle distances collected by each of the Y-axis infrared sensors at the time point and the Y-axis obstacle distances collected by each of the Y-axis infrared sensors at the previous time point, calculating the Y-axis change parameter at the time point; for each of the time points, based on the X-axis obstacle distances collected by each of the X-axis infrared sensors at the time point and the X-axis obstacle distances collected by each of the X-axis infrared sensors at the previous time point, calculating the X-axis change parameter at the time point; for each of the time points, weightedly summing the Y-axis change parameter at the time point and the X-axis change parameter at the time point to obtain the vehicle posture change parameter of the vehicle to be identified at the time point.

[0009] Furthermore, the parameter determination unit determines the vehicle model parameters of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point, including: determining the maximum vehicle length and maximum vehicle height of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the three-dimensional coordinate information of each Y-axis infrared sensor; determining the maximum vehicle width of the vehicle to be identified based on the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point and the three-dimensional coordinate information of each X-axis infrared sensor; determining the vehicle chassis height of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the three-dimensional coordinate information of each Y-axis infrared sensor, as well as the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point and the three-dimensional coordinate information of each X-axis infrared sensor.

[0010] Furthermore, the parameter determination unit predicts the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle model parameters of the vehicle to be identified, including: obtaining the vehicle model parameters and relative positions of the license plates of multiple test vehicles, wherein the relative positions of the license plates are used to characterize the relative positions of the center of the license plate of the test vehicle and the center points of the two front wheels of the test vehicle; clustering the multiple test vehicles based on the vehicle model parameters of the multiple test vehicles to determine multiple vehicle clusters; based on the Y-axis obstacle distance collected by each of the Y-axis infrared sensors at the target time point and the three-dimensional coordinate information of each of the Y-axis infrared sensors and the X-axis infrared sensors at the target time point The three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified are determined based on the X-axis obstacle distance collected by the point and the three-dimensional coordinate information of each X-axis infrared sensor; a similar vehicle cluster is determined from the multiple vehicle clusters based on the vehicle model parameters of the vehicle to be identified; a similar test vehicle is determined from the similar vehicle cluster based on the vehicle model parameters of the vehicle to be identified; the three-dimensional coordinates of the license plate of the vehicle to be identified are predicted based on the relative position of the license plate of the similar test vehicle and the three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified, wherein the three-dimensional coordinates of the license plate of the vehicle to be identified include the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

[0011] Furthermore, the image acquisition module determines whether to adjust the image acquisition posture based on the predicted license plate position of the vehicle to be identified, including: obtaining multiple first training samples, wherein the first training samples include sample image acquisition postures, images acquired by the image acquisition device under the sample image acquisition postures, and three-dimensional coordinates of multiple markers, and the labels of the first training samples are pixel coordinates of multiple markers; establishing a pixel coordinate prediction model; training the pixel coordinate prediction model through the multiple first training samples; using the trained pixel coordinate prediction model, based on the current image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified, predicting the first pixel coordinates of the four corners of the license plate of the vehicle to be identified; and judging whether to adjust the image acquisition posture based on the first pixel coordinates of the four corners of the license plate of the vehicle to be identified.

[0012] Furthermore, the image acquisition module generates a target image acquisition pose based on the license plate position of the vehicle to be identified, including: obtaining multiple second training samples, wherein the second training samples include the sample image acquisition pose before adjustment and the three-dimensional coordinates of multiple markers, and the label of the second training sample is the sample image acquisition pose after adjustment; establishing a pose adjustment model; training the pose adjustment model through the multiple second training samples; and using the trained pose adjustment model to generate the target image acquisition pose based on the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

[0013] Furthermore, the license plate recognition module identifies the license plate information of the vehicle to be identified based on the image of the vehicle to be identified, including: predicting the second pixel coordinates of the four corners of the license plate of the vehicle to be identified based on the target image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified through the pixel coordinate prediction model; extracting the target area image from the image of the vehicle to be identified based on the predicted second pixel coordinates of the four corners of the license plate of the vehicle to be identified; and identifying the license plate information of the vehicle to be identified based on the target area image.

[0014] Furthermore, the license plate recognition module identifies the license plate information of the vehicle to be identified based on the target area image, including: extracting a background area image and multiple license plate character images based on the target area image; judging whether the license plate of the vehicle to be identified is genuine based on the background area image and multiple license plate character images; when it is determined that the license plate of the vehicle to be identified is genuine, performing optical character recognition on the multiple license plate character images to determine the license plate number of the vehicle to be identified; when it is determined that the license plate of the vehicle to be identified is fake, generating a warning message.

[0015] Compared with the existing technology, the vehicle gate automatic license plate recognition control system provided in this specification has at least the following beneficial effects:

[0016] 1. Multiple X-axis infrared sensors positioned along the width of the barrier's railings and multiple Y-axis infrared sensors along its length precisely capture infrared information from vehicles, enabling more accurate prediction of the three-dimensional coordinates of the license plate. Based on these predicted coordinates, the system intelligently determines whether the image capture device's position needs to be adjusted. If the current position is not conducive to clearly capturing the license plate image, the system automatically adjusts to the optimal position, ensuring high-quality images every time, thereby improving image capture efficiency and success rate. The system is adaptable to vehicles of varying sizes, types, and driving postures. Thanks to its multi-dimensional infrared sensing and intelligent position adjustment mechanism, it maintains high recognition rates and stability even in complex and changing traffic scenarios. This automated and intelligent license plate recognition process reduces manual intervention and waiting time, improving vehicle traffic efficiency. For drivers, this means shorter wait times and a smoother traffic experience. By reducing manual intervention and improving recognition efficiency, the system can significantly reduce operating costs for parking lots and related management agencies. Furthermore, high-quality image capture and recognition reduces disputes and losses caused by misidentification or missed recognition.

[0017] 2. By analyzing the changing posture parameters of the vehicle at multiple consecutive time points, the system can more accurately grasp the vehicle's dynamic driving state. Using the Y-axis obstacle distances collected by each Y-axis infrared sensor and the X-axis obstacle distances collected by each X-axis infrared sensor at the target time, the system determines not only the vehicle's maximum width, length, and height, but also its chassis height. This comprehensive vehicle model identification helps the system better understand the characteristics of different vehicles, providing a more accurate parameter basis for subsequent license plate recognition. By performing cluster analysis on the vehicle model parameters and relative license plate positions of multiple test vehicles, the system learns the relationship between different vehicle models and license plate positions. This enables the system to more quickly identify similar vehicle clusters and similar test vehicles across different makes and models, thereby improving the system's adaptability and robustness. After determining the three-dimensional coordinates of the license plate, the system can perform more targeted image acquisition and processing. This avoids unnecessary image acquisition and processing, conserves image processing resources, and improves overall system efficiency and performance.

[0018] 3. The system intelligently determines whether the image acquisition device's pose needs to be adjusted based on the predicted license plate position of the vehicle to be identified. This dynamic adjustment ensures that each license plate image captured is clear and unobstructed, thereby improving image acquisition quality. Using a trained pixel coordinate prediction model, the system can quickly predict the license plate's pixel coordinates within the image and determine whether the image acquisition pose needs to be adjusted. This rapid prediction and judgment reduces unnecessary image acquisition and adjustment time, improving image acquisition efficiency. The system can adapt to variations in license plate position and angle for different vehicles. By training the pixel coordinate prediction model and pose adjustment model, the system learns the relationship between the pixel coordinates and image acquisition pose of different vehicle license plates, enabling rapid adjustments to ensure accurate license plate recognition. Existing barrier gate image acquisition methods use a fixed image acquisition angle, which can result in incomplete license plate images. By introducing the pixel coordinate prediction model and pose adjustment model, the system can more precisely control the image acquisition process, reducing unnecessary resource consumption.

[0019] 4. The system first uses a pixel coordinate prediction model to predict the precise location of the license plate within the image. It then extracts an image of the target area based on this location. This approach reduces the impact of background noise and other interfering factors on license plate recognition, thereby improving recognition accuracy. Before recognizing the license plate information, the system also verifies the authenticity of the license plate. By extracting images of the background area and the license plate characters, the system can analyze characteristics such as the integrity, color, and font of the license plate, effectively identifying forged or tampered license plates. This anti-counterfeiting capability is crucial for maintaining traffic order and public safety. When recognizing a license plate, the system processes only the extracted target area image, rather than analyzing the entire vehicle image. This approach reduces the amount of data processed, thereby increasing processing speed and enabling vehicles to pass through the barrier more quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, like numbers represent like structures, wherein:

[0021] Figure 1 This is a module diagram of a vehicle barrier automatic license plate recognition control system shown in one embodiment of the present application;

[0022] Figure 2 This is a flowchart of predicting the three-dimensional coordinates of the license plate of a vehicle to be identified, shown in one embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly introduces the drawings required for describing the embodiments.

[0024] Figure 1 This is a module diagram of a vehicle gate automatic license plate recognition control system shown in one embodiment of the present application, such as Figure 1 As shown, a vehicle barrier automatic license plate recognition control system may include a vehicle sensing module, a vehicle recognition module, an image acquisition module and a license plate recognition module.

[0025] The vehicle sensing module can be used to sense whether there is a vehicle in the lane.

[0026] Specifically, the vehicle sensing module may include a ground sensing coil, which can be installed under the ground of the barrier. When a vehicle passes by, the metal part of the vehicle will induce the ground sensing coil, thereby changing the induced current in the coil. By detecting the change in the induced current, it can be determined whether there is a vehicle in the lane.

[0027] The vehicle identification module may include an infrared sensing unit and a parameter determination unit, wherein the infrared sensing unit includes a plurality of X-axis infrared sensing devices arranged along the width direction of the barrier and a plurality of Y-axis infrared sensing devices arranged along the length direction of the barrier, the X-axis infrared sensing device includes a plurality of X-axis infrared sensors arranged along the vertical direction, and the Y-axis infrared sensing device includes a plurality of Y-axis infrared sensors arranged along the vertical direction. The infrared sensing unit is used to collect vehicle infrared sensing information when the vehicle sensing module senses that there is a vehicle in the lane, and the parameter determination unit is used to predict the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle infrared sensing information.

[0028] Specifically, multiple Y-axis infrared sensors are arranged along a line parallel to the length of the barrier. As a vehicle travels along the barrier, the multiple Y-axis infrared sensors sequentially detect the presence of the vehicle as it moves. The infrared light emitted by the X-axis infrared sensor is perpendicular to the length of the barrier, while the infrared light emitted by the Y-axis infrared sensor is perpendicular to the infrared light emitted by the X-axis infrared sensor. The infrared light emitted by any two Y-axis infrared sensors is parallel to each other.

[0029] Furthermore, when the vehicle sensing module senses that there is a vehicle in the lane, the infrared sensing unit collects vehicle infrared sensing information, including:

[0030] Start each Y-axis infrared sensor and collect the Y-axis obstacle distance, where the Y-axis obstacle distance is used to represent the straight-line distance between the side of the vehicle and the Y-axis infrared sensor;

[0031] Each X-axis infrared sensor is activated to collect the X-axis obstacle distance, wherein the vehicle infrared sensing information includes at least the Y-axis obstacle distance and the X-axis obstacle distance. The X-axis obstacle distance is used to represent the straight-line distance between the front of the vehicle and the X-axis infrared sensor.

[0032] Furthermore, the parameter determination unit predicts the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle infrared sensing information, including:

[0033] Determine vehicle posture change parameters of the vehicle to be identified at multiple consecutive time points based on the Y-axis obstacle distances collected by each Y-axis infrared sensor at multiple consecutive time points and the X-axis obstacle distances collected by each X-axis infrared sensor at multiple consecutive time points;

[0034] Determining a target time point from the multiple consecutive time points based on vehicle posture change parameters of the vehicle to be identified at the multiple consecutive time points;

[0035] Determining the vehicle type parameters of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point, wherein the vehicle type parameters of the vehicle to be identified include at least the maximum width, maximum length, maximum height, and chassis height of the vehicle;

[0036] Based on the model parameters of the vehicle to be identified, the three-dimensional coordinates of the license plate of the vehicle to be identified are predicted.

[0037] Furthermore, the parameter determination unit determines the vehicle posture change parameters of the vehicle to be identified at the multiple consecutive time points based on the Y-axis obstacle distances collected by each Y-axis infrared sensor at the multiple consecutive time points and the X-axis obstacle distances collected by each X-axis infrared sensor at the multiple consecutive time points, including:

[0038] For each time point, based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the time point and the Y-axis obstacle distance collected by each Y-axis infrared sensor at the previous time point, calculate the Y-axis change parameter at the time point;

[0039] For each time point, based on the X-axis obstacle distance collected by each X-axis infrared sensor at the time point and the X-axis obstacle distance collected by each X-axis infrared sensor at the previous time point, calculate the X-axis change parameter at the time point;

[0040] For each time point, the Y-axis change parameter and the X-axis change parameter of the time point are weighted and summed to obtain the vehicle posture change parameter of the vehicle to be identified at the time point.

[0041] The Y-axis change parameter at a time point can be calculated according to the following formula:

[0042]

[0043] Among them, V (t,Y) is the Y-axis change parameter at the t-th time point, a i is the weight corresponding to the i-th Y-axis infrared sensing device, a i is greater than 0, and V (t,i) is the Y-axis change parameter of the i-th Y-axis infrared sensing device at the t-th time point, I is the total number of Y-axis infrared sensing devices, D ((i,j,t),Y) D is the Y-axis obstacle distance collected by the j-th Y-axis infrared sensor included in the i-th Y-axis infrared sensing device at the t-th time point, ((i,j,t-1),Y) is the Y-axis obstacle distance collected by the j-th Y-axis infrared sensor included in the i-th Y-axis infrared sensing device at the t-1-th time point (i.e., the time point before the t-th time point), and J is the total number of Y-axis infrared sensors included in the i-th Y-axis infrared sensing device.

[0044] The calculation method of the X-axis change parameter of the time point is similar to the calculation method of the Y-axis change parameter of the time point, and will not be repeated here.

[0045] The vehicle posture change parameters of the vehicle to be identified at a certain time point can be calculated according to the following formula:

[0046] V t =b1×V (t,Y) +b2×V (t,X)

[0047] Among them, V t is the vehicle posture change parameter of the vehicle to be identified at the tth time point, b1 and b2 are weights, and b1 and b2 are greater than 0, b1+b2=1, V (t,X) is the X-axis variation parameter at the t-th time point.

[0048] If there is a time point at which the vehicle posture change parameter is less than the vehicle posture change parameter threshold, and the vehicle posture change parameters at time points before this time point are all greater than or equal to the vehicle posture change parameter threshold, and the vehicle posture change parameters at time points after this time point are all less than the vehicle posture change parameter threshold, then this time point is selected as the target time point. It can be understood that at the target time point, the vehicle to be identified is in a state of approximately stopping.

[0049] Furthermore, the parameter determination unit determines the vehicle type parameters of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point, including:

[0050] Determine the maximum length and maximum height of the vehicle to be identified based on the Y-axis obstacle distance collected by each Y-axis infrared sensor at the target time point and the three-dimensional coordinate information of each Y-axis infrared sensor;

[0051] Determine the maximum width of the vehicle to be identified based on the X-axis obstacle distance collected by each X-axis infrared sensor at the target time point and the three-dimensional coordinate information of each X-axis infrared sensor;

[0052] The vehicle chassis height of the vehicle to be identified is determined based on the Y-axis obstacle distance and three-dimensional coordinate information of each Y-axis infrared sensor collected at the target time point by each Y-axis infrared sensor, and the X-axis obstacle distance and three-dimensional coordinate information of each X-axis infrared sensor collected at the target time point by each X-axis infrared sensor.

[0053] Specifically, for each Y-axis infrared sensing device, if there is a Y-axis infrared sensor among the multiple Y-axis infrared sensors included in the Y-axis infrared sensing device that has a Y-axis infrared sensor quantity greater than a preset Y-axis infrared sensor quantity threshold and the Y-axis obstacle distance collected at the target time point is less than the Y-axis obstacle distance threshold, then the Y-axis infrared sensing device is regarded as the target Y-axis infrared sensing device.

[0054] For each target Y-axis infrared sensing device, the Y-axis infrared sensor among the multiple Y-axis infrared sensors included in the target Y-axis infrared sensing device whose Y-axis obstacle distance, collected at the target time, is less than the Y-axis obstacle distance threshold is used as the target Y-axis infrared sensor included in the target Y-axis infrared sensing device. Based on the three-dimensional coordinate information of the target Y-axis infrared sensors, the height of the highest target Y-axis infrared sensor can be used as the vehicle height of the vehicle to be identified corresponding to the target Y-axis infrared sensing device.

[0055] The maximum value of the vehicle heights of the vehicle to be identified corresponding to each target Y-axis infrared sensing device is taken as the maximum vehicle height of the vehicle to be identified.

[0056] For any two target Y-axis infrared sensing devices, calculate the distance between the target Y-axis infrared sensors located at the same height in the two target Y-axis infrared sensing devices. For example, if target Y-axis infrared sensing device A1 and target Y-axis infrared sensing device A2 are located, and target Y-axis infrared sensing device A1 includes target Y-axis infrared sensor B1, and target Y-axis infrared sensing device A2 includes target Y-axis infrared sensor B2 located at the same height as target Y-axis infrared sensor B1, calculate the distance between target Y-axis infrared sensor B1 and target Y-axis infrared sensor B2. The maximum distance between the two target Y-axis infrared sensors located at the same height is taken as the maximum vehicle length of the vehicle to be identified.

[0057] For each X-axis infrared sensing device, if there is an X-axis infrared sensor among the multiple X-axis infrared sensors included in the X-axis infrared sensing device that has an X-axis infrared sensor number greater than a preset X-axis infrared sensor number threshold and the X-axis obstacle distance collected at the target time point is less than the X-axis obstacle distance threshold, then the X-axis infrared sensing device is regarded as the target X-axis infrared sensing device.

[0058] For each target X-axis infrared sensing device, the X-axis infrared sensor whose X-axis obstacle distance, collected at the target time, is less than the Y-axis obstacle distance threshold among the multiple X-axis infrared sensors included in the target X-axis infrared sensing device is used as the target X-axis infrared sensor included in the target X-axis infrared sensing device. Based on the three-dimensional coordinate information of the target X-axis infrared sensor, the straight-line distance between the two target X-axis infrared sensors at both ends can be calculated as the vehicle width of the target vehicle corresponding to the target X-axis infrared sensing device.

[0059] The maximum vehicle width of the vehicle to be identified corresponding to each target X-axis infrared sensing device is taken as the maximum vehicle width of the vehicle to be identified.

[0060] The vehicle chassis height of the vehicle to be identified can be determined by a chassis height prediction model based on the Y-axis obstacle distance collected by each target Y-axis infrared sensor at a target time point and the three-dimensional coordinate information of each target Y-axis infrared sensor, as well as the X-axis obstacle distance collected by each target X-axis infrared sensor at a target time point and the three-dimensional coordinate information of each target X-axis infrared sensor. The chassis height prediction model can be a convolutional neural network (CNN) model.

[0061] Figure 2 This is a flow chart showing the prediction of the three-dimensional coordinates of the license plate of a vehicle to be identified in one embodiment of the present application. Figure 2 As shown, further, the parameter determination unit predicts the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle model parameters of the vehicle to be identified, including:

[0062] Obtaining vehicle model parameters and license plate relative positions of multiple test vehicles, wherein the license plate relative position is used to represent the relative position between the license plate center of the test vehicle and the center points of the two front wheels of the test vehicle;

[0063] Clustering the multiple test vehicles based on their vehicle model parameters to determine multiple vehicle clusters;

[0064] Determine the three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified based on the Y-axis obstacle distance and three-dimensional coordinate information of each Y-axis infrared sensor collected at the target time point, and the X-axis obstacle distance and three-dimensional coordinate information of each X-axis infrared sensor collected at the target time point;

[0065] Determine similar vehicle clusters from multiple vehicle clusters based on vehicle model parameters of the vehicle to be identified;

[0066] Based on the model parameters of the vehicle to be identified, a similar test vehicle is determined from a cluster of similar vehicles;

[0067] Based on the relative positions of the license plates of similar test vehicles and the three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified, the three-dimensional coordinates of the license plate of the vehicle to be identified are predicted, wherein the three-dimensional coordinates of the license plate of the vehicle to be identified include the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

[0068] Specifically, for any two test vehicles, the vehicle model similarity of the two test vehicles can be calculated based on the vehicle model parameters of the two test vehicles. A clustering algorithm (for example, a K-means clustering algorithm, a hierarchical clustering algorithm, etc.) can be used to cluster multiple test vehicles based on the vehicle model similarity of any two test vehicles to determine multiple vehicle clusters.

[0069] The model similarity of two test vehicles can be calculated based on the following formula:

[0070]

[0071] Among them, S (e,f) is the similarity of the two test vehicles, c1, c2, c3 and c4 are weights, c1, c2, c3 and c4 are greater than 0, c1+c2+c3+c4=1, P1, P2, P3 and P4 are preset parameters, P1, P2, P3 and P4 are greater than 0, W e is the maximum width of the e-th test vehicle, W f is the maximum width of the f-th test vehicle, L e is the maximum length of the e-th test vehicle, L f is the maximum length of the f-th test vehicle, H (e,1) is the maximum height of the e-th test vehicle, H (f,1) is the maximum height of the f-th test vehicle, H (e,2) is the chassis height of the e-th test vehicle, H (f,2) is the chassis height of the f-th test vehicle.

[0072] The three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified can be determined by a first coordinate prediction model based on the Y-axis obstacle distance collected by each target Y-axis infrared sensor at the target time point and the three-dimensional coordinate information of each target Y-axis infrared sensor, as well as the X-axis obstacle distance collected by each target X-axis infrared sensor at the target time point and the three-dimensional coordinate information of each target X-axis infrared sensor. The first coordinate prediction model can be a generative adversarial network (GAN) model.

[0073] The vehicle model similarity between the vehicle to be identified and the cluster center of the similar vehicle cluster can be calculated based on the vehicle model parameters of the vehicle to be identified and the vehicle model parameters corresponding to the cluster center of the similar vehicle cluster. The method of calculating the vehicle model similarity between the vehicle to be identified and the cluster center of the similar vehicle cluster is similar to the method of calculating the vehicle model similarity between two test vehicles, which will not be repeated here.

[0074] Vehicle clusters whose vehicle model similarity exceeds a first vehicle model similarity threshold are considered similar vehicle clusters. Based on the vehicle model parameters of the vehicle to be identified and the vehicle model parameters of each test vehicle included in the similar vehicle cluster, the vehicle model similarity between the vehicle to be identified and each test vehicle included in the similar vehicle cluster is calculated. Test vehicles included in the similar vehicle cluster whose vehicle model similarity exceeds a second vehicle model similarity threshold are considered similar vehicles.

[0075] The three-dimensional coordinates of the license plate of the vehicle to be identified are predicted based on the relative positions of the license plates of similar test vehicles and the three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified through a second coordinate model, wherein the second coordinate model can be a generative adversarial network model.

[0076] The image acquisition module can be used to determine whether to adjust the image acquisition posture based on the predicted three-dimensional coordinates of the license plate of the vehicle to be identified. If it is determined that the image acquisition posture adjustment is required, the target image acquisition posture is generated based on the three-dimensional coordinates of the license plate of the vehicle to be identified, and the posture of the image acquisition device is adjusted based on the target image acquisition posture. After the posture is adjusted, the image of the vehicle to be identified is acquired. If it is determined that the image acquisition posture adjustment is not required, the image of the vehicle to be identified is acquired.

[0077] Furthermore, the image acquisition module determines whether to adjust the image acquisition posture based on the predicted license plate position of the vehicle to be identified, including:

[0078] Acquire a plurality of first training samples, wherein the first training samples include a sample image acquisition posture, an image acquired by an image acquisition device in the sample image acquisition posture, and three-dimensional coordinates of a plurality of markers, and labels of the first training samples are pixel coordinates of the plurality of markers;

[0079] Establishing a pixel coordinate prediction model, wherein the pixel coordinate prediction model may be a generative adversarial network model;

[0080] Training a pixel coordinate prediction model using a plurality of first training samples;

[0081] Using the trained pixel coordinate prediction model, based on the current image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified, predict the first pixel coordinates of the four corners of the license plate of the vehicle to be identified;

[0082] Based on the first pixel coordinates of the four corners of the license plate of the vehicle to be identified, it is determined whether to adjust the image acquisition posture.

[0083] Specifically, if the first pixel coordinates of at least one corner exceed the pixel coordinate range corresponding to the current image acquisition pose, a decision is made to adjust the image acquisition pose. If the first pixel coordinates of all four corners of the license plate of the vehicle to be identified are within the pixel coordinate range corresponding to the current image acquisition pose, a decision is made not to adjust the image acquisition pose.

[0084] Furthermore, the image acquisition module generates a target image acquisition pose based on the license plate position of the vehicle to be identified, including:

[0085] Acquire multiple second training samples, wherein the second training samples include the sample image acquisition pose before adjustment and the three-dimensional coordinates of multiple markers, and the labels of the second training samples are the sample image acquisition pose after adjustment;

[0086] Establishing a posture adjustment model, wherein the posture adjustment model can be a generative adversarial network model;

[0087] Training a pose adjustment model using multiple second training samples;

[0088] Using the trained pose adjustment model, the target image acquisition pose is generated based on the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

[0089] The license plate recognition module can be used to identify the license plate information of the vehicle to be identified based on the image of the vehicle to be identified.

[0090] Specifically include:

[0091] Predicting the second pixel coordinates of the four corners of the license plate of the vehicle to be identified based on the target image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified by the pixel coordinate prediction model;

[0092] extracting a target area image from the image of the vehicle to be identified based on the predicted second pixel coordinates of the four corners of the license plate of the vehicle to be identified;

[0093] Based on the target area image, the license plate information of the vehicle to be identified is identified.

[0094] Specifically, the target area image may be extracted from the image of the vehicle to be identified by using the predicted second pixel coordinates of the four corners of the license plate of the vehicle to be identified as the pixel coordinates of the four corners of the target area image.

[0095] Furthermore, the license plate recognition module identifies the license plate information of the vehicle to be identified based on the target area image, including:

[0096] Based on the target area image, extract the background area image and multiple license plate character images;

[0097] Based on the background area image and the multiple license plate character images, determining whether the license plate of the vehicle to be identified is genuine;

[0098] When it is determined that the license plate of the vehicle to be identified is genuine, optical character recognition is performed on the multiple license plate character images to determine the license plate number of the vehicle to be identified;

[0099] When the license plate of the vehicle to be identified is determined to be fake, a warning message is generated.

[0100] Specifically, genuine license plates have unique font designs, a pleasant visual experience, and uniform font size and structure. Counterfeit license plates may be too thin, too thick, or unusually shaped, or have irregularities in the font curves. Furthermore, the system can detect any traces of erasure, obscuration, or pasting on the license plate, such as white or black self-adhesive fonts pasted onto the plate, or hollowed-out or cut-out fonts from genuine plates to patch together.

[0101] Authentic license plates are produced using a one-step molding process, resulting in a uniform overall shape with neat corners, arcs, and other details. Counterfeit license plates may have varying shapes and sizes. Authentic license plates have high-quality paint, with even coloring applied only to the protruding parts. Counterfeit license plates may have poor-quality paint and uneven coloring.

[0102] Based on the above-mentioned distinction between genuine and fake license plates, a license plate authenticity recognition model can be used to determine whether the license plate of the vehicle to be identified is genuine based on the background area image and multiple license plate character images, wherein the license plate authenticity recognition model can be a convolutional neural network model.

[0103] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A vehicle gate automatic license plate recognition control system, characterized in that: include: Vehicle sensing module, used to sense whether there is a vehicle in the lane; A vehicle identification module includes an infrared sensing unit and a parameter determination unit, wherein the infrared sensing unit includes a plurality of X-axis infrared sensing devices arranged along the width direction of the barrier railing and a plurality of Y-axis infrared sensing devices arranged along the length direction of the barrier, the X-axis infrared sensing device includes a plurality of X-axis infrared sensors arranged along the length direction of the barrier railing, and the Y-axis infrared sensing device includes a plurality of Y-axis infrared sensors arranged along the vertical direction, the infrared sensing unit is used to collect vehicle infrared sensing information when the vehicle sensing module senses that there is a vehicle in the lane, and the parameter determination unit is used to predict the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle infrared sensing information; An image acquisition module is configured to determine whether to adjust the image acquisition posture based on the predicted three-dimensional coordinates of the license plate of the vehicle to be identified; if it is determined that the image acquisition posture adjustment is to be performed, a target image acquisition posture is generated based on the three-dimensional coordinates of the license plate of the vehicle to be identified; the posture of the image acquisition device is adjusted based on the target image acquisition posture; after the posture adjustment, an image of the vehicle to be identified is acquired; if it is determined that the image acquisition posture adjustment is not to be performed, an image of the vehicle to be identified is acquired; a license plate recognition module, configured to recognize the license plate information of the vehicle to be recognized based on the image of the vehicle to be recognized; Determining the vehicle chassis height of the vehicle to be identified based on the Y-axis obstacle distance and three-dimensional coordinate information of each Y-axis infrared sensor collected at the target time point by each Y-axis infrared sensor, and the X-axis obstacle distance and three-dimensional coordinate information of each X-axis infrared sensor collected at the target time point by each X-axis infrared sensor; and predicting the three-dimensional coordinates of the license plate of the vehicle to be identified based on the vehicle model parameters of the vehicle to be identified by the parameter determination unit, including: Obtaining vehicle model parameters and license plate relative positions of multiple test vehicles, wherein the license plate relative positions are used to represent the relative positions of the license plate centers of the test vehicles and the center points of the two front wheels of the test vehicles; Clustering the multiple test vehicles based on vehicle model parameters of the multiple test vehicles to determine multiple vehicle clusters; Determining a similar test vehicle from a cluster of similar vehicles based on the vehicle model parameters of the vehicle to be identified; Based on the relative position of the license plate of the similar test vehicle and the three-dimensional coordinates of the center points of the two front wheels of the vehicle to be identified, the three-dimensional coordinates of the license plate of the vehicle to be identified are predicted, wherein the three-dimensional coordinates of the license plate of the vehicle to be identified include the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

2. A vehicle barrier automatic license plate recognition control system according to claim 1, characterized in that: The image acquisition module determines whether to adjust the image acquisition posture based on the predicted license plate position of the vehicle to be identified, including: Acquire a plurality of first training samples, wherein the first training samples include a sample image acquisition posture, an image acquired by an image acquisition device in the sample image acquisition posture, and three-dimensional coordinates of a plurality of markers, and labels of the first training samples are pixel coordinates of the plurality of markers; Establish a pixel coordinate prediction model; Training the pixel coordinate prediction model using the multiple first training samples; Using the trained pixel coordinate prediction model, based on the current image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified, predict the first pixel coordinates of the four corners of the license plate of the vehicle to be identified; Based on the first pixel coordinates of the four corners of the license plate of the vehicle to be identified, it is determined whether to adjust the image acquisition posture.

3. A vehicle barrier automatic license plate recognition control system according to claim 2, characterized in that: The image acquisition module generates a target image acquisition posture based on the license plate position of the vehicle to be identified, including: Acquire a plurality of second training samples, wherein the second training samples include a sample image acquisition pose before adjustment and three-dimensional coordinates of a plurality of markers, and labels of the second training samples are the sample image acquisition pose after adjustment; Establish a posture adjustment model; Training the posture adjustment model using the plurality of second training samples; The trained posture adjustment model is used to generate the target image acquisition posture based on the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified.

4. A vehicle barrier automatic license plate recognition control system according to claim 2 or 3, characterized in that: The license plate recognition module recognizes the license plate information of the vehicle to be recognized based on the image of the vehicle to be recognized, including: Predicting second pixel coordinates of the four corners of the license plate of the vehicle to be identified by the pixel coordinate prediction model based on the target image acquisition posture and the three-dimensional coordinates of the four corners of the license plate of the vehicle to be identified; extracting a target area image from the image of the vehicle to be identified based on the predicted second pixel coordinates of the four corners of the license plate of the vehicle to be identified; Based on the target area image, the license plate information of the vehicle to be identified is identified.

5. The vehicle gate automatic license plate recognition control system according to claim 4 is characterized in that: The license plate recognition module recognizes the license plate information of the vehicle to be recognized based on the target area image, including: Extracting a background area image and a plurality of license plate character images based on the target area image; Based on the background area image and the plurality of license plate character images, determining whether the license plate of the vehicle to be identified is genuine; When it is determined that the license plate of the vehicle to be identified is genuine, performing optical character recognition on the multiple license plate character images to determine the license plate number of the vehicle to be identified; When it is determined that the license plate of the vehicle to be identified is fake, a warning message is generated.

Images