Road parking inspection system and method based on unmanned aerial vehicle

Abstract

The invention discloses a road parking inspection system and method based on an unmanned aerial vehicle, and relates to the field of intelligent parking. The system comprises an autonomous navigation module which carries out autonomous navigation according to a preset inspection route; the image acquisition module is used for acquiring a parking space image through a camera; the image recognition module processes and analyzes the image data according to an image recognition algorithm; the mobile communication module transmits an identification result to the intelligent parking management platform through a mobile communication technology; and the intelligent parking management platform judges the vehicle violation condition according to the recognition result. The road parking inspection system provided by the invention is high in detection precision and flexible in detection area, realizes unmanned management of roadside parking spaces, can guarantee the timeliness of road inspection, and improves the efficiency of road inspection.

Description

Technical Field

[0001] The present invention relates to the field of smart parking, and in particular to a road parking inspection system and method based on an unmanned aerial vehicle. Background Art

[0002] In recent years, cars have become an indispensable means of transportation for people's daily travel, and the problem of parking difficulties has become increasingly prominent, especially in busy commercial areas, residential areas and near public transportation hubs. It is very easy to cause collisions and even lead to some serious traffic accidents. At present, most inspections for illegal parking of vehicles are still carried out by manual inspections. During the inspection process, the inspectors record the license plates of illegally parked vehicles and take photos for evidence. This method is not only a waste of human resources but also has relatively low inspection efficiency. When the number of illegally parked vehicles is large, manual inspections are prone to recording errors. Therefore, the development of a simple and efficient road parking inspection method is imminent.

[0003] Most of the existing road parking inspection methods currently have defects in some aspects. For example, the road parking inspection method based on video inspection vehicles generally uses a small vehicle as a carrier, installs a camera on the inspection vehicle to shoot illegally parked vehicles, and transmits the license plate and parking space information of the illegally parked vehicles to the operation management platform after the shooting is completed; the video inspection vehicle used in this method needs to travel on specific urban roads, and its inspection efficiency is easily affected by the current road traffic conditions. During peak traffic hours, the inspection efficiency of the video inspection vehicle will be greatly reduced due to road congestion, and it will also affect the normal driving of other vehicles on the road. The inspection efficiency is difficult to guarantee, and the video inspection vehicle needs to be manually driven, and it has not achieved full automation. Therefore, based on this problem, this application proposes a road parking inspection system and method based on an unmanned aerial vehicle. Summary of the invention

[0004] Technical Purpose In order to solve the above problems, the purpose of the present invention is to provide a road parking inspection system and method based on an unmanned aerial vehicle, which not only solves the problem of time-consuming and labor-intensive traditional manual inspections, but also avoids the occurrence of inaccurate inspection results due to human factors, realizes the simplification and automation of road inspections, improves the inspection accuracy, inspection efficiency and adaptability to different inspection scenarios of road inspections, and avoids the problem of low inspection efficiency due to environmental factors.

[0005] Technical Solution In order to achieve the above-mentioned purpose, the present invention provides a road parking inspection system and method based on an unmanned aerial vehicle, which adopts an unmanned aerial vehicle as a carrier, realizes the autonomous navigation function of the inspection unmanned aerial vehicle through a GPS positioning device and a navigation algorithm, realizes the automatic obstacle avoidance function during the autonomous navigation process by carrying an infrared sensor, and processes and analyzes the vehicle parking space image through an image recognition algorithm, so as to monitor the parking situation of vehicles in the parking area in real time.

[0006] In a first aspect, the present invention provides a road parking inspection system based on an unmanned aerial vehicle, comprising: Autonomous navigation module, which performs autonomous navigation according to the preset inspection route and realizes automatic obstacle avoidance during autonomous navigation; An image acquisition module, which acquires a parking space image through a camera, wherein the parking space image includes a vehicle license plate number and a parking space number; An image recognition module processes and analyzes image data according to an image recognition algorithm to obtain license plate information and parking space information, and determines the parking space corresponding to the captured image by the time of the picture frame, the time corresponding to the license plate shooting, and the position of the parking space number in the image; wherein the image recognition algorithm includes an OCR text recognition algorithm; Mobile communication module, which transmits the recognition results to the smart parking management platform through mobile communication technology; The smart parking management platform determines vehicle violations based on the recognition results.

[0007] The system body is a patrol unmanned aerial vehicle, and the autonomous navigation module, the image acquisition module, the image recognition module and the mobile communication module are integrated into the patrol unmanned aerial vehicle. The patrol unmanned aerial vehicle can be connected with other smart parking systems to achieve data sharing and collaborative work between systems.

[0008] Furthermore, the autonomous navigation module realizes the autonomous navigation function of the inspection unmanned aerial vehicle according to the GPS positioning device and the navigation algorithm, and realizes the automatic obstacle avoidance function during the autonomous navigation process by carrying an infrared sensor, and the automatic obstacle avoidance function is specifically to automatically adjust the flight route when an obstacle is detected on the flight route. The autonomous navigation module can improve the adaptability of the inspection unmanned aerial vehicle to complex environments, enable it to work effectively in a variety of scenarios, avoid the impact of road traffic conditions on inspection efficiency, and further realize automation.

[0009] Furthermore, the image acquisition module captures berth images through a camera in the following steps: obtaining area ranges and area distributions of a plurality of parking areas, and generating a route plan covering each parking area according to the area ranges and the area distributions; inputting the route plan into an unmanned aerial vehicle inspection system, so that the inspection unmanned aerial vehicle performs autonomous navigation according to the route plan, and performs image capture on each parking area to obtain the berth image.

[0010] Furthermore, after receiving the video stream collected by the berth number acquisition camera, the image recognition module obtains the berth number data through the OCR text recognition algorithm built into the image recognition module. It obtains the berth number data by preprocessing the picture frames in the video stream, segmenting the berth number characters in the picture frames, and identifying the content of the berth number characters; the OCR text recognition algorithm supports a variety of models, including traditional template matching algorithms, support vector machines, Bayesian classification algorithms, and neural network models based on deep learning. It can select a suitable model for training according to specific application scenarios and needs; after the model training is completed, the image recognition module improves the accuracy of the OCR text recognition algorithm through data set optimization, feature selection, and model structure optimization. It not only solves the time-consuming and labor-intensive problems of traditional manual inspections, but also avoids the occurrence of inaccurate inspection results due to human factors; it realizes the simplification and automation of road inspections and improves the inspection accuracy of unmanned aerial vehicles.

[0011] Furthermore, after receiving the identification result transmitted by the inspection unmanned aerial vehicle, the smart parking management platform pre-judges the current geographical location of the unmanned aerial vehicle. If it is determined that the current geographical location of the unmanned aerial vehicle is not within the area of ​​the parking area, it is determined that the current vehicle is illegally parked. If the geographical location information is consistent, the received identification result is uploaded to the parking space database of the platform for matching. If the matching results of the license plate information and the parking space information are inconsistent, it is determined that the current vehicle occupies the parking space.

[0012] Furthermore, the smart parking management platform pre-stores a parking space atlas of each parking area, which lists each parking space in the corresponding parking area in the form of points; after receiving the parking space image, the smart parking management platform converts each vehicle on the parking space image into a point through a built-in algorithm, and then compares and analyzes it with the parking space atlas of the parking area to find that some parking spaces are not parked, and some points are not parking spaces but have vehicles parked, which can be identified as illegal parking. Through the above scheme, during the communication between the unmanned aerial vehicle and the smart parking management platform, it can be determined whether there is illegal parking and the remaining space information of the parking area can be obtained, and whether there is illegal parking, and then the parking area can be monitored and supervised in real time to effectively prevent illegal parking. It can also display the remaining space information of the parking area to the car owner by accessing the public network, guide the car owner to enter the parking area, and improve the use efficiency and economic benefits of the parking area.

[0013] Furthermore, the system unifies the path planning problem into a multi-objective optimization model by constructing an inspection map and removing redundant information. It quickly covers the predetermined inspection area through optimized path planning, improves the execution speed of the inspection task, reduces the round-trip time and distance of the UAV, and reduces the operating cost; the multi-objective optimization model ensures the balanced use of the UAVs, avoiding the problem of some UAVs being overused while others are idle.

[0014] Furthermore, the system implements the function of the inspection UAV to quickly identify the area where the license plate is located in the image after obtaining the parking space image based on the Yolo algorithm solution, so that the OCR algorithm model can better adapt to target detection under different angles and lighting conditions, and provide accurate location information for subsequent OCR recognition. It improves the detection accuracy of road parking inspections, enhances the recognition ability of UAV image extraction features, and can accurately identify and distinguish different types of parking states; it reduces model parameters and calculations, so that the inspection UAV can achieve rapid target detection with limited computing resources.

[0015] In a second aspect, the present invention further provides a road parking inspection method based on an unmanned aerial vehicle, the method being based on the system described in the first aspect, comprising: Autonomous navigation according to preset inspection routes; Collect berth images through cameras; Process and analyze image data according to image recognition algorithms; Transmitting identification results via mobile communication technology; The vehicle violation is determined based on the recognition results.

[0016] In a third aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the aforementioned road parking inspection method based on an unmanned aerial vehicle is implemented.

[0017] The present invention uses an unmanned aerial vehicle as a carrier, realizes the autonomous navigation function of the inspection unmanned aerial vehicle through a GPS positioning device and a navigation algorithm, realizes the automatic obstacle avoidance function during the autonomous navigation process by carrying an infrared sensor, and processes and analyzes the vehicle parking image through an image recognition algorithm; by constructing an inspection map and removing redundant information, the path planning problem is unified into a multi-objective optimization model; based on the Yolo algorithm scheme, the inspection unmanned aerial vehicle can quickly identify the area where the license plate is located in the image after obtaining the parking image. The system and method not only solve the problem of time-consuming and labor-intensive traditional manual inspections, but also avoid the occurrence of inaccurate inspection results caused by human factors, realize the simplification and automation of road inspections, improve the inspection accuracy, inspection efficiency and adaptability to different inspection scenes of road inspections, and avoid the problem of low inspection efficiency due to environmental factors.

[0018] Beneficial Effects By implementing the above-mentioned unmanned aerial vehicle-based road parking inspection system and method provided by the present invention, the following technical effects are achieved: (1) The present invention uses an unmanned aerial vehicle as a carrier, realizes the autonomous navigation function of the inspection unmanned aerial vehicle through a GPS positioning device and a navigation algorithm, and realizes the automatic obstacle avoidance function during autonomous navigation by carrying an infrared sensor; it has high adaptability to complex environments and can work effectively in a variety of scenarios, avoiding the impact of road traffic conditions on inspection efficiency and further realizing automation; it adopts a modular design and can be connected to other smart parking systems to realize data sharing and collaborative work between systems.

[0019] (2) The present invention processes and analyzes vehicle parking space images through an image recognition algorithm; it not only solves the time-consuming and labor-intensive problem of traditional manual inspections, but also avoids the occurrence of inaccurate inspection results due to human factors; it simplifies and automates road inspections and improves the inspection accuracy of unmanned aerial vehicles.

[0020] (3) The present invention unifies the path planning problem into a multi-objective optimization model by constructing an inspection map and removing redundant information; it quickly covers the predetermined inspection area through optimized path planning, improves the execution speed of the inspection task, reduces the time and distance of the UAV's round trip, and reduces the operating cost; the multi-objective optimization model ensures the balanced use of the UAVs, avoiding the problem of overuse of some UAVs while other UAVs are idle.

[0021] (4) The present invention realizes the function of the inspection unmanned aerial vehicle to quickly identify the area where the license plate is located in the image after obtaining the parking space image based on the Yolo algorithm scheme; it improves the detection accuracy of road parking inspection, enhances the recognition ability of unmanned aerial vehicle image extraction features, and can accurately identify and distinguish different types of parking status; it reduces model parameters and calculation amount, so that the inspection unmanned aerial vehicle can achieve rapid target detection with limited computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to make the above-mentioned unmanned aerial vehicle-based road parking inspection system and method of the present invention more obvious and easy to understand, the following is a brief introduction to the drawings required for use in the specific implementation methods of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 A schematic diagram showing steps of a road parking inspection method; Figure 2 A schematic diagram showing the comparison of parking space atlases (A is the parking space atlas of the parking area, and B is the parking space atlas of the parking area); Figure 3 A schematic diagram of an inspection map is shown. DETAILED DESCRIPTION

[0024] Embodiment 1: A road parking inspection system based on an unmanned aerial vehicle is provided, the system comprising: an autonomous navigation module, which performs autonomous navigation according to a preset inspection route and realizes automatic obstacle avoidance during the autonomous navigation process; an image acquisition module, which acquires a parking space image through a camera, wherein the parking space image includes a vehicle license plate number and a parking space number; an image recognition module, which processes and analyzes image data according to an image recognition algorithm to obtain license plate information and parking space information, and determines the parking space corresponding to the captured image through the time of the picture frame, the time corresponding to the license plate shooting, and the position of the parking space number in the image; wherein the image recognition algorithm includes an OCR text recognition algorithm; and a mobile communication module, which transmits the recognition result to a smart parking management platform through mobile communication technology.

[0025] The road parking inspection system based on unmanned aerial vehicle provided by the present invention specifically includes: The system body is a patrol unmanned aerial vehicle, and the autonomous navigation module, the image acquisition module, the image recognition module and the mobile communication module are integrated into the patrol unmanned aerial vehicle. The patrol unmanned aerial vehicle can be connected with other smart parking systems to achieve data sharing and collaborative work between systems.

[0026] The autonomous navigation module realizes the autonomous navigation function of the inspection unmanned aerial vehicle based on the GPS positioning device and the navigation algorithm, and realizes the automatic obstacle avoidance function during the autonomous navigation process by carrying an infrared sensor. The automatic obstacle avoidance function specifically automatically adjusts the flight route when an obstacle is detected on the flight route.

[0027] The steps of the image acquisition module acquiring parking space images through a camera include: acquiring the area range and area distribution of multiple parking areas, and generating a route plan covering each parking area according to the area range and the area distribution; inputting the route plan into the unmanned aerial vehicle inspection system, so that the inspection unmanned aerial vehicle performs autonomous navigation according to the route plan, and acquires images of each parking area to obtain the parking space image.

[0028] The vehicle's license plate number is captured and identified in real time by an intelligent license plate recognition camera installed on the vehicle inspection vehicle; the berth number is captured and identified in real time by a special camera installed on the vehicle inspection vehicle for collecting a clear berth number video stream.

[0029] After receiving the video stream captured by the berth number acquisition camera, the image recognition module obtains the berth number data through the OCR text recognition algorithm built into the image recognition module. It obtains the berth number data by preprocessing the picture frames in the video stream, segmenting the berth number characters in the picture frames, and identifying the content of the berth number characters; the OCR text recognition algorithm supports multiple models, including traditional template matching algorithms, support vector machines, Bayesian classification algorithms, and neural network models based on deep learning. It can select appropriate models for training according to specific application scenarios and needs; after the model training is completed, the image recognition module improves the accuracy of the OCR text recognition algorithm through data set optimization, feature selection, and model structure optimization.

[0030] After receiving the identification result transmitted by the inspection unmanned aerial vehicle, the smart parking management platform pre-judges the current geographical location of the unmanned aerial vehicle. If it is determined that the current geographical location of the unmanned aerial vehicle is not within the area of ​​the parking area, it is determined that the current vehicle is illegally parked. If the geographical location information is consistent, the received identification result is uploaded to the parking space database of the platform for matching. If the matching results of the license plate information and the parking space information are inconsistent, it is determined that the current vehicle occupies the parking space.

[0031] The smart parking management platform pre-stores a collection of parking space locations in each parking area, such as Figure 2 As shown, each parking space is arranged in the form of points in the corresponding parking area, such as Figure 2As shown in A; after receiving the parking space image, the smart parking management platform converts each vehicle on the parking space image into a point through a built-in algorithm to obtain a parking point atlas of the parking area as shown in Figure 2 As shown in B, then by comparing and analyzing with the parking space atlas of the parking area, it can be seen that some parking spaces are not parked, and some points are not parking spaces but have vehicles parked, which can be identified as illegal parking. Through the above scheme, during the communication between the unmanned aerial vehicle and the smart parking management platform, it can be determined whether there is illegal parking and the remaining space information of the parking area can be obtained. Then, the parking area can be monitored and supervised in real time to effectively prevent illegal parking. It can also display the remaining space information of the parking area to the car owner by accessing the public network, guide the car owner to enter the parking lot, and improve the use efficiency and economic benefits of the parking area.

[0032] The system unifies the path planning problem into a multi-objective optimization model by constructing an inspection map and removing redundant information, including: constructing an inspection map and removing redundant information irrelevant to road inspection to simplify the model; unifying the path planning problem into a multi-objective optimization model; optimizing route planning using particle swarm optimization, and redesigning the encoding strategy, update rules and decoding strategy of particles; the inspection map is shown in the figure below. Figure 3 shown.

[0033] Embodiment 2: On the basis of the above-mentioned embodiment, a Yolo algorithm solution is added to realize the function of the inspection unmanned aerial vehicle to quickly identify the area where the license plate is located in the image after obtaining the parking space image, so that the OCR algorithm model can better adapt to target detection under different angles and lighting conditions, and provide accurate location information for subsequent OCR recognition; Among them, it is necessary to preprocess the original image in advance and perform data enhancement on the original image:

[0034] in, Represents an image after data augmentation; represents the original image; Represents a set of transformations, including random rotation, scaling, cropping, color adjustment, etc.

[0035] Assume that the Backbone network is a deep convolutional neural network, which can be represented as a combination of a series of convolution and activation functions:

[0036] in, Represents the output feature map of a specific layer; Represents a layer of the BackBone network; represents the activation function; Indicates The weight of the layer; Indicates The bias of the layer.

[0037] The feature pyramid is constructed through the PAFPN structure, and feature maps of different levels are fused. Multi-scale feature representation is constructed through upsampling and feature fusion:

[0038] in, Indicates PAFPN output of the layer; Represents Backbone's Layer output; Indicates from Mapping of layers to higher layers; Represents an upsampling operation; Represents the feature fusion operation.

[0039] Apply depth-wise separable convolution to the output feature map of PAFPN to reduce the amount of computation, including depth-wise convolution and point-wise convolution:

[0040]

[0041] in, represents the depth convolution output; Indicates A deep filter; Indicates the number of input channels; Represents the feature map obtained by deep convolution; Represents the point-wise convolution output.

[0042] Enhance the contextual information of feature maps through dilated convolution and feature fusion:

[0043] in, represents the convolution kernel size, represents the expansion rate; Represents the feature map after deep context enhancement; Represents a deep context augmentation operation.

[0044] Spatial filtering and channel filtering:

[0045]

[0046]

[0047] in, represents the spatial filtering operation; represents the spatial filtering output; Represents channel filtering operation; Indicates the channel filter output; Represents the operation of connecting feature maps in the channel dimension; Represents the feature map after concatenation in the channel dimension.

[0048] The object detection head includes classification, localization and object prediction:

[0049]

[0050]

[0051] in, Represents the weight of the detection head; Indicates the offset of the detection head; Represents the activation function.

[0052] The loss function includes classification loss, positioning loss and target loss:

[0053]

[0054]

[0055]

[0056] in, represents the binary cross entropy loss; represents the bounding box regression loss; represents the loss weight.

[0057] Positive and negative sample matching is usually based on the IoU threshold:

[0058] in, Represents the real labeled data.

[0059] Post-processing:

[0060]

[0061] in, Indicates the number of candidate detection boxes before non-maximum suppression; Represents the evaluation indicator.

[0062] For example, there is a dataset that contains 100 images of road parking taken by unmanned aerial vehicles. Each image contains a different number of parking targets. The image size is , the target vehicle size ranges from Pixel to Pixels are not equal; Original Image Size , apply a random rotation of 5 degrees, a translation of plus or minus 2 pixels, and a brightness change of plus or minus 10%; Feature extraction formula:

[0063] Assume that the first layer output for ; Through the PAFPN structure, we get feature maps of three scales, the sizes are , and ; Applying depthwise separable convolution, assuming the input The number of channels is 32 and the number of output channels is 64:

[0064] Use dilated convolution with a dilation rate of 2 to enhance contextual information:

[0065] Assume that spatial filtering uses The maximum pooling and channel filtering are used The detection head predicts 3 categories, and each feature point corresponds to 4 bounding boxes; the real annotation Contains bounding boxes and categories of 20 vehicles; the IoU threshold is 0.5, matching 15 positive samples and 5 negative samples; filtering is applied with a threshold of 0.3 and ; Apply NMS to get the final detection result:

[0066]

[0067] After training, the mAP of the model on the validation set reached 0.75, indicating that the model has good detection accuracy.

[0068] The effect of the Yolo algorithm solution is shown in Table 1: Table 1. Summary of Yolo algorithm solution effects

[0069] As shown in the table, the application of the Yolo algorithm greatly improves the performance of the recognition module. Before the application of the algorithm, the detection accuracy of the system was 0.65, and after the application, it increased to 0.82, achieving a gain of 25.8%, which shows that the Yolo algorithm can more accurately identify parking targets; at the same time, the precision rate increased from 0.70 to 0.85, and the recall rate increased from 0.60 to 0.80, reducing the occurrence of false alarms, further proving that the model has made progress in both precision and recall; in addition, the Yolo algorithm also shortened the average detection time from 1.2 seconds to 0.5 seconds, meeting the needs of real-time processing and improving detection efficiency.

[0070] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable non-transient storage media containing computer-usable program code.

[0071] The present invention can provide computer program instructions to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the system.

[0072] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions of the system.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions of the system.

Claims

1. A road parking inspection system based on an unmanned aerial vehicle, characterized in that: include: Autonomous navigation module, which performs autonomous navigation according to the preset inspection route and realizes automatic obstacle avoidance during autonomous navigation; An image acquisition module, which acquires a parking space image through a camera, wherein the parking space image includes a vehicle license plate number and a parking space number; An image recognition module processes and analyzes image data according to an image recognition algorithm to obtain license plate information and parking space information, and determines the parking space corresponding to the captured image by the time of the picture frame, the time corresponding to the license plate shooting, and the position of the parking space number in the image; wherein the image recognition algorithm includes an OCR text recognition algorithm; Mobile communication module, which transmits the recognition results to the smart parking management platform through mobile communication technology; Smart parking management platform, judging vehicle violations based on recognition results; Inspection map module, used to unify the path planning problem into a multi-objective optimization model; The Yolo processing module is used to enable the inspection UAV to quickly identify the area where the license plate is located in the image after obtaining the parking space image.

2. The road parking inspection system based on unmanned aerial vehicle according to claim 1 is characterized by: The system body is a patrol unmanned aerial vehicle, and the autonomous navigation module, the image acquisition module, the image recognition module and the mobile communication module are integrated into the patrol unmanned aerial vehicle. The patrol unmanned aerial vehicle can be connected with other smart parking systems to achieve data sharing and collaborative work between systems.

3. The road parking inspection system based on unmanned aerial vehicle according to claim 1 or 2, characterized in that: The autonomous navigation module realizes the autonomous navigation function of the inspection unmanned aerial vehicle based on the GPS positioning device and the navigation algorithm, and realizes the automatic obstacle avoidance function during the autonomous navigation process by carrying an infrared sensor. The automatic obstacle avoidance function specifically automatically adjusts the flight route when an obstacle is detected on the flight route.

4. The road parking inspection system based on unmanned aerial vehicle according to claim 1 or 2, characterized in that: The step of obtaining the berth image includes: obtaining the area range and area distribution of multiple parking areas, and generating a route plan covering each parking area according to the area range and the area distribution; inputting the route plan into the unmanned aerial vehicle inspection system, so that the inspection unmanned aerial vehicle performs autonomous navigation according to the route plan, and collects images of each parking area to obtain the berth image.

5. The road parking inspection system based on unmanned aerial vehicle according to claim 1 is characterized in that: After receiving the video stream captured by the berth number acquisition camera, the image recognition module obtains the berth number data through the OCR text recognition algorithm built into the image recognition module. It obtains the berth number data by preprocessing the picture frames in the video stream, segmenting the berth number characters in the picture frames, and identifying the content of the berth number characters; the OCR text recognition algorithm supports multiple models, including traditional template matching algorithms, support vector machines, Bayesian classification algorithms, and neural network models based on deep learning. It can select appropriate models for training according to specific application scenarios and needs; after the model training is completed, the image recognition module improves the accuracy of the OCR text recognition algorithm through data set optimization, feature selection, and model structure optimization.

6. The road parking inspection system based on unmanned aerial vehicle according to claim 1 or 2, characterized in that: After receiving the identification result transmitted by the inspection unmanned aerial vehicle, the smart parking management platform pre-judges the current geographical location of the unmanned aerial vehicle. If it is determined that the current geographical location of the unmanned aerial vehicle is not within the area of ​​the parking area, it is determined that the current vehicle is illegally parked. If the geographical location information is consistent, the received identification result is uploaded to the parking space database of the platform for matching. If the matching results of the license plate information and the parking space information are inconsistent, it is determined that the current vehicle occupies the parking space.

7. The road parking inspection system based on unmanned aerial vehicle according to claim 1 is characterized in that: The Yolo processing module uses a deep convolutional neural network to enable the OCR algorithm model to better adapt to target detection under different angles and lighting conditions, providing accurate position information for OCR recognition.

8. A road parking inspection method based on an unmanned aerial vehicle, characterized in that: The method is implemented based on the system according to any one of claims 1 to 7: The method comprises: Autonomous navigation according to preset inspection routes; Collect berth images through cameras; Process and analyze image data according to image recognition algorithms; Transmitting identification results via mobile communication technology; The vehicle violation is determined based on the recognition results.

9. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, characterized in that: When the computer program is executed, the road parking inspection method based on the unmanned aerial vehicle described in claim 8 is implemented.

Images