A smart parking lot safety management method
By laying cameras and using recognition models in smart parking lots, unattended vehicle monitoring and alarms are achieved, the problems of low efficiency and high cost of traditional parking lot safety management models are solved, the intelligent level of parking lot safety management is improved, and the reduction of Disputes and theft risks.
Patent Information
- Application Number
- CN202411320549.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-09-23
AI Technical Summary
The traditional car park safety management model has problems such as low efficiency, high cost, inability to fully cover the car park, the inability to track the vehicle's driving trajectory alone, difficulty in identifying unlicensed vehicles and small vehicles, and inadequate detection of abnormal events is not smart enough.
Camera is installed in smart parking lots, using vehicle identification models, collision identification models and robbery identification models, and through V2X communication, the vehicle identity, motion trajectory and video stream are monitored in real time, and collision and robbery events are identified and alarmed to achieve unattended Safety management.
It realizes vehicle monitoring with full coverage, quickly identify and alarm collisions and robbery incidents, reduces the cost of manual patrol, improves the intelligent level of parking safety management, promptly detects and reports abnormal situations, and reduces the risks of disputes and robbery.
Smart Images

Figure CN118840874B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a traffic control system, and in particular to an intelligent parking lot safety management method. Background Art
[0002] With the rapid development of social economy, the number of parking lots is increasing, and parking lot safety management is facing more and more severe challenges. There are two main traditional parking lot safety management modes: one is the traditional manual patrol. The other is to add on-site monitoring, unified access to the monitoring room, and staff on duty to monitor abnormalities. Both methods have many problems and shortcomings.
[0003] Traditional manual patrols are inefficient and difficult to fully cover every corner of the parking lot. They are prone to security loopholes and are not enough to effectively respond to increasingly complex security needs. Secondly, the high labor cost increases the operating cost of parking lot management. The monitoring system in the monitoring room can often only provide a separate video stream and cannot accurately track the driving trajectory of the vehicle across the video. At the same time, there are difficulties such as identifying unlicensed vehicles and accurately identifying small vehicles. Simple monitoring is not smart enough in detecting abnormal events. It is often necessary to manually check the monitoring screen to find problems, and the response speed is slow. In addition, the existing parking lot safety management solutions cannot be effectively linked with the vehicle's own safety system. When abnormal situations such as vehicle collisions occur, the parking lot management system cannot obtain accurate information in time and take corresponding measures.
[0004] Sentry mode is a safety feature built for smart connected cars based on the vehicle's existing sensors, cameras and other hardware. It helps the driver to obtain real-time information about the vehicle's safety when leaving the vehicle, and automatically collects and records images around the vehicle and sends an alert to the driver's mobile phone. However, it can currently only record on-site conditions and send prompts to car owners. Summary of the invention
[0005] The purpose of the present invention is to provide an intelligent parking lot safety management method that solves the above-mentioned problems, improves the parking lot safety management level, improves the intelligence level, and reduces labor costs.
[0006] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is as follows: a smart parking lot safety management method is applied to a smart parking lot equipped with a management terminal, and the management terminal can communicate with the vehicle system of the vehicle through V2X, and the smart parking lot is equipped with a camera for obtaining vehicle images, vehicle monitoring images and monitoring videos, comprising the following steps:
[0007] S1, deploying vehicle identification model, vehicle collision identification model and vehicle theft identification model on the management end;
[0008] The vehicle recognition model is used to extract vehicle features from vehicle images, and the vehicles include licensed vehicles and unlicensed vehicles;
[0009] The vehicle collision recognition model is used to recognize collision events based on vehicle monitoring images collected in the smart parking lot, and the collision events include vehicle-to-vehicle collisions and human-to-vehicle collisions;
[0010] The vehicle theft identification model is used to identify whether a vehicle has been stolen or not based on vehicle monitoring images collected in the smart parking lot;
[0011] S2, the vehicle drives to the entrance gate, obtains a vehicle image as a first vehicle image, identifies identity information from the first vehicle image and stores it in a vehicle identity table, wherein the identity information includes vehicle features, vehicle body color and driver facial information;
[0012] S3, when a vehicle moves in the smart parking lot, it is subjected to mobile monitoring, wherein the mobile monitoring is to locate the vehicle in real time, generate a motion trajectory and store it in the vehicle identity table, and obtain the vehicle monitoring image in real time to identify the collision event. If a collision event is identified, a first alarm signal is sent to the management end;
[0013] S4, when a vehicle is parked in the smart parking lot for more than a preset time T1, parking monitoring is performed on it, wherein the parking monitoring is to obtain the location of the parking place and match the parking space number;
[0014] If the parking place is not in the parking space, it is identified as abnormal parking, and a second alarm signal is sent to the management end;
[0015] If the vehicle is parked in a parking space, it is recognized as normal parking and the parking space number is stored in the vehicle identity table;
[0016] S5: From the time the vehicle is parked normally to the time it is started again, the vehicle computer turns on the sentry mode, and the smart parking lot monitors the vehicle in real time to identify whether a collision or theft occurs;
[0017] If the vehicle triggers an alarm in sentry mode, a third alarm signal is sent to the management terminal;
[0018] If a collision or theft event is identified, a fourth alarm signal is sent to the management end;
[0019] S6: When the vehicle is identified as being parked normally and when the vehicle reaches the exit gate, a vehicle identity comparison is performed respectively, wherein:
[0020] When the vehicle is identified as being parked normally, the camera corresponding to the parking space number in the vehicle identity table is called to obtain a vehicle image as a second vehicle image, from which identity information is identified and compared with the identity information in the vehicle identity table. If at least one item is different, a fifth alarm signal is sent to the management end;
[0021] When the vehicle reaches the exit gate, the vehicle identity table corresponding to the vehicle is determined according to the vehicle trajectory, and the vehicle image is obtained at the exit gate as the third vehicle image, from which the identity information is identified. If at least one item is different, a sixth alarm signal is sent to the management end.
[0022] Preferably, in S1, the method for extracting vehicle features using the vehicle recognition model comprises steps S11 to S12.
[0023] S11, obtaining a vehicle image of the vehicle, and sending it to a license plate recognition model to identify the license plate number. If the license plate number is recognized, the license plate number is used as the vehicle feature of the vehicle. If the license plate number is not recognized, the vehicle image is sent to an unlicensed vehicle recognition model;
[0024] S12, using the unlicensed vehicle recognition model to recognize the vehicle image, if a small car is recognized therefrom, using the CILP model to extract the image features of the vehicle image as vehicle features, if the small car is not recognized, then the vehicle will not be allowed to pass, the small car includes two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles whose size is smaller than a preset size.
[0025] Preferably, in S1, the unlicensed vehicle recognition model is trained by a YOLOv8 network based on a small vehicle dataset D1, wherein the samples in the small vehicle dataset D1 are manually labeled vehicle images of two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles whose sizes are smaller than a preset size.
[0026] Preferably, the vehicle collision recognition model is obtained by training a YOLOv8 network according to a collision data set D2, wherein the samples in the collision data set D2 are manually labeled images of collisions between people and vehicles and images of collisions between vehicles;
[0027] The vehicle theft recognition model is obtained by training the YOLOv8 network based on the theft dataset D3, where the samples in the theft dataset D3 are manually marked theft images;
[0028] The vehicle body color is extracted from the vehicle image by a vehicle body color recognition model.
[0029] As a preferred embodiment: S3 further includes, when the management terminal receives the first alarm signal, selecting the vehicle monitoring image and the location of the vehicle determined as a collision event, and sending them to the administrator;
[0030] S4 also includes, when the management end receives the second alarm signal, obtaining the location of the vehicle and sending it to the administrator.
[0031] As a preferred embodiment: S5 also includes, when the management end receives the third alarm signal or the fourth alarm signal, reading the parking space number from the vehicle identity table, obtaining the surveillance videos of 1-3 cameras closest to the parking space number, each video captures a video clip of T2 minutes before the alarm signal, and sending the parking space number and video clip to the administrator.
[0032] As a preference: S6 also includes, when the management terminal receives the fifth alarm signal or the sixth alarm signal, saving the second vehicle image or the third vehicle image, and querying whether there is owner information corresponding to the vehicle feature in the management terminal, and notifying the owner if there is.
[0033] Compared with the prior art, the advantages of the present invention are:
[0034] (1) The present invention deploys cameras covering the entire parking lot in the smart parking lot to obtain relevant vehicle images, vehicle monitoring images and monitoring videos. When a vehicle enters the parking lot, parks normally, or exits the parking lot, the relevant cameras obtain three vehicle images, extract identity information respectively, and then compare the second and third identity information with the first identity information. If the abnormality described in the present invention is found, the relevant data is retained by storing the second and third vehicle images or notifying the owner, which is convenient for subsequent verification. It can play an anti-theft effect to a certain extent.
[0035] (2) The present invention uses a vehicle recognition model to identify licensed and unlicensed vehicles, extracts unique vehicle features of each vehicle for identity recognition, uses a vehicle collision recognition model to realize unattended collision event recognition, and uses a vehicle theft recognition model to realize unattended theft event recognition. After a vehicle enters the smart parking lot, whether it is moving or parked, there is a corresponding monitoring program to monitor it, which can quickly and effectively identify anomalies such as collision, theft, and parking in a non-parking space, and promptly notify the administrator for manual intervention.
[0036] (3) After the vehicle enters the smart parking lot and parks normally, the sentinel mode is turned on and communication with the management end is maintained. In this way, even if the smart parking lot does not detect any abnormality, the sentinel mode can detect the abnormality and promptly notify the administrator for manual intervention.
[0037] In summary, the present invention can monitor vehicles from the entire process of entering to leaving the parking lot, can replace manual patrols, actively discover and promptly report the location of incidents, and effectively reduce the possibility of disputes or thefts in the parking lot. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0039] The present invention will be further described below in conjunction with the accompanying drawings.
[0040] Example 1: See Figure 1A smart parking lot safety management method is applied to a smart parking lot equipped with a management terminal capable of performing V2X communication with a vehicle system, wherein the smart parking lot is equipped with cameras for acquiring vehicle images, vehicle monitoring images and monitoring videos, and comprises the following steps:
[0041] S1, deploying vehicle identification model, vehicle collision identification model and vehicle theft identification model on the management end;
[0042] The vehicle recognition model is used to extract vehicle features from vehicle images, and the vehicles include licensed vehicles and unlicensed vehicles;
[0043] The vehicle collision recognition model is used to recognize collision events based on vehicle monitoring images collected in the smart parking lot, and the collision events include vehicle-to-vehicle collisions and human-to-vehicle collisions;
[0044] The vehicle theft identification model is used to identify whether a vehicle has been stolen or not based on vehicle monitoring images collected in the smart parking lot;
[0045] S2, the vehicle drives to the entrance gate, obtains a vehicle image as a first vehicle image, identifies identity information from the first vehicle image and stores it in a vehicle identity table, wherein the identity information includes vehicle features, vehicle body color and driver facial information;
[0046] S3, when a vehicle moves in the smart parking lot, it is subjected to mobile monitoring, wherein the mobile monitoring is to locate the vehicle in real time, generate a motion trajectory and store it in the vehicle identity table, and obtain the vehicle monitoring image in real time to identify the collision event. If a collision event is identified, a first alarm signal is sent to the management end;
[0047] S4, when a vehicle is parked in the smart parking lot for more than a preset time T1, parking monitoring is performed on it, wherein the parking monitoring is to obtain the location of the parking place and match the parking space number;
[0048] If the parking place is not in the parking space, it is identified as abnormal parking, and a second alarm signal is sent to the management end;
[0049] If the vehicle is parked in a parking space, it is recognized as normal parking and the parking space number is stored in the vehicle identity table;
[0050] S5: From the time the vehicle is parked normally to the time it is started again, the vehicle computer turns on the sentry mode, and the smart parking lot monitors the vehicle in real time to identify whether a collision or theft occurs;
[0051] If the vehicle triggers an alarm in sentry mode, a third alarm signal is sent to the management terminal;
[0052] If a collision or theft event is identified, a fourth alarm signal is sent to the management end;
[0053] S6: When the vehicle is identified as being parked normally and when the vehicle reaches the exit gate, a vehicle identity comparison is performed respectively, wherein:
[0054] When the vehicle is identified as being parked normally, the camera corresponding to the parking space number in the vehicle identity table is called to obtain a vehicle image as a second vehicle image, from which identity information is identified and compared with the identity information in the vehicle identity table. If at least one item is different, a fifth alarm signal is sent to the management end;
[0055] When the vehicle reaches the exit gate, the vehicle identity table corresponding to the vehicle is determined according to the vehicle trajectory, and the vehicle image is obtained at the exit gate as the third vehicle image, from which the identity information is identified. If at least one item is different, a sixth alarm signal is sent to the management end.
[0056] In S1, the method for extracting vehicle features using a vehicle recognition model includes steps S11 to S12.
[0057] S11, obtaining a vehicle image of the vehicle, and sending it to a license plate recognition model to identify the license plate number. If the license plate number is recognized, the license plate number is used as the vehicle feature of the vehicle. If the license plate number is not recognized, the vehicle image is sent to an unlicensed vehicle recognition model;
[0058] S12, using the unlicensed vehicle recognition model to recognize the vehicle image, if a small car is recognized therefrom, using the CILP model to extract the image features of the vehicle image as vehicle features, if the small car is not recognized, then the vehicle will not be allowed to pass, the small car includes two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles whose size is smaller than a preset size.
[0059] In S1, the unlicensed vehicle recognition model is trained by the YOLOv8 network based on the small vehicle dataset D1. The samples in the small vehicle dataset D1 are manually labeled vehicle images of two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles whose sizes are smaller than a preset size.
[0060] The vehicle collision recognition model is trained by the YOLOv8 network based on the collision data set D2, where the samples in the collision data set D2 are manually labeled human-vehicle collision pictures and vehicle-vehicle collision pictures;
[0061] The vehicle theft recognition model is obtained by training the YOLOv8 network based on the theft dataset D3, where the samples in the theft dataset D3 are manually marked theft images;
[0062] The vehicle body color is extracted from the vehicle image by a vehicle body color recognition model.
[0063] In the present invention, the vehicle recognition model is to extract unique vehicle features from the vehicle image and match it with the vehicle identity. Therefore, for licensed vehicles, the license plate recognition model is used to identify the license plate number. There are many models based on different license plate recognition algorithms on the market. The present invention does not improve them. Any one can be selected to identify the license plate number from the vehicle image. For unlicensed vehicles, the present invention uses the unlicensed vehicle recognition model CILP model combined to extract vehicle features. First, the purpose of the unlicensed vehicle recognition model is that when the vehicle passes through the entrance gate, the camera at the entrance gate collects the vehicle image, and then identifies small vehicles such as two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles with a size smaller than the preset size from the vehicle image. If a small vehicle is identified, it means that there is a target in the vehicle image, and then this vehicle image is sent to the CILP model to extract the image features of the vehicle image as vehicle features. The CILP model does not need to be trained and is only used to extract vehicle features. If the small vehicle is not identified, it means that there is no target in the vehicle image, and there is no need to use the CILP model for feature extraction. After the vehicle recognition model, whether it is a licensed vehicle or an unlicensed vehicle, a unique vehicle feature is obtained.
[0064] The unlicensed vehicle recognition model is obtained by training the YOLOv8 network based on the small vehicle data set D1. The construction method of the small vehicle data set D1 is to obtain a large number of images containing two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles with a size smaller than a preset size, and take the two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles with a size smaller than a preset size as targets, mark the bounding boxes and categories of the targets in the images, and form samples. All samples constitute the small vehicle data set D1. Then, D1 is divided into a training set, a validation set, and a test set. The YOLOv8 network is selected, and the parameters such as the learning rate, batch size, number of iterations, optimization algorithm, and loss function are set. Then, the divided data set is input into the set YOLOv8 network, and the prediction is performed through forward propagation, and the optimization is optimized through back propagation. It is iterated until the stopping condition is met to obtain the unlicensed vehicle recognition model. The stopping condition may include the loss value converging to below a certain threshold, reaching a preset number of iterations, and the performance on the validation set is no longer significantly improved.
[0065] The vehicle collision recognition model is trained by the YOLOv8 network based on the collision dataset D2. When constructing the collision dataset D2, a large number of images of human-vehicle collisions and vehicle-vehicle collisions in different scenarios are collected and manually labeled to form the samples of D2.
[0066] The vehicle theft and robbery recognition model is trained by the YOLOv8 network based on the theft and robbery dataset D3, and the samples in the theft and robbery dataset D3 are also manually labeled.
[0067] There are many ways to implement the vehicle color recognition model. For example, in a vehicle color recognition method, the target detection algorithm obtains multiple candidate areas from the vehicle image, and then selects a certain area from the candidate areas to extract its color features. Another example is a vehicle color recognition system, which determines the vehicle body area after collecting the vehicle image, uses an image processing device to calculate the color distribution map of the vehicle body position area in the RGB color space, and counts the color with the largest distribution area in the color distribution map as the vehicle body color.
[0068] The driver's facial information is actually obtained by using a face recognition algorithm. At the entrance gate, the camera captures video, from which the clearest frame of image is selected for face recognition.
[0069] For mobile monitoring, combined with the cameras at the entrance gate, lanes, and parking spaces, the vehicle can be located in real time to generate a motion trajectory. Later, when the vehicle leaves the site, it is also a moving process and a motion trajectory will also be generated. When the vehicle moves to the exit gate, it will first analyze which vehicle it is based on the motion trajectory, and then use a third vehicle image for identity recognition.
[0070] Parking monitoring is actually the monitoring and management of whether the vehicle is parked normally. In addition, in the present invention, when the vehicle is parked normally and is started again, the sentinel mode will be turned on and communicate with the management end in real time, so that the management end can combine the information of the sentinel mode for safety management.
[0071] Example 2: See Figure 1 Based on Example 1, we make the following improvements:
[0072] S3 also includes, when the management terminal receives the first alarm signal, selecting the vehicle monitoring image determined as a collision event and the location of the vehicle, and sending them to the administrator.
[0073] S4 also includes, when the management end receives the second alarm signal, obtaining the location of the vehicle and sending it to the administrator.
[0074] S5 also includes, when the management end receives the third alarm signal or the fourth alarm signal, reading the parking space number from the vehicle identity table, obtaining the surveillance videos of 1-3 cameras closest to the parking space number, each video captures a video clip of T2 minutes before the alarm signal, and sending the parking space number and video clip to the administrator.
[0075] S6 also includes, when the management terminal receives the fifth alarm signal or the sixth alarm signal, saving the second vehicle image or the third vehicle image, and querying whether there is vehicle owner information corresponding to the vehicle feature in the management terminal, and notifying the vehicle owner if there is.
[0076] The rest is the same as Example 1.
[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A smart parking lot safety management method is applied to a smart parking lot equipped with a management terminal capable of performing V2X communication with a vehicle's vehicle-to-vehicle system, wherein the smart parking lot is equipped with cameras for acquiring vehicle images, vehicle monitoring images, and monitoring videos, and is characterized in that: The following steps are included: S1, deploying vehicle identification model, vehicle collision identification model and vehicle theft identification model on the management end; The vehicle recognition model is used to extract vehicle features from vehicle images, and the vehicles include licensed vehicles and unlicensed vehicles; The vehicle collision recognition model is used to recognize collision events based on vehicle monitoring images collected in the smart parking lot, and the collision events include vehicle-to-vehicle collisions and human-to-vehicle collisions; The vehicle theft identification model is used to identify whether a vehicle has been stolen or not based on vehicle monitoring images collected in the smart parking lot; S2, the vehicle drives to the entrance gate, obtains a vehicle image as a first vehicle image, identifies identity information from the first vehicle image and stores it in a vehicle identity table, wherein the identity information includes vehicle features, vehicle body color and driver facial information; S3, when a vehicle moves in the smart parking lot, it is subjected to mobile monitoring, wherein the mobile monitoring is to locate the vehicle in real time, generate a motion trajectory and store it in the vehicle identity table, and obtain the vehicle monitoring image in real time to identify the collision event. If a collision event is identified, a first alarm signal is sent to the management end; S4, when a vehicle is parked in the smart parking lot for more than a preset time T1, parking monitoring is performed on it, wherein the parking monitoring is to obtain the location of the parking place and match the parking space number; If the parking place is not in the parking space, it is identified as abnormal parking, and a second alarm signal is sent to the management end; If the vehicle is parked in a parking space, it is recognized as normal parking and the parking space number is stored in the vehicle identity table; S5: From the time the vehicle is parked normally to the time it is started again, the vehicle computer turns on the sentry mode, and the smart parking lot monitors the vehicle in real time to identify whether a collision or theft occurs; If the vehicle triggers an alarm in sentry mode, a third alarm signal is sent to the management terminal; If a collision or theft event is identified, a fourth alarm signal is sent to the management end; S6: When the vehicle is identified as being parked normally and when the vehicle reaches the exit gate, a vehicle identity comparison is performed respectively, wherein: When the vehicle is identified as being parked normally, the camera corresponding to the parking space number in the vehicle identity table is called to obtain a vehicle image as a second vehicle image, from which identity information is identified and compared with the identity information in the vehicle identity table. If at least one item is different, a fifth alarm signal is sent to the management end; When the vehicle reaches the exit gate, the vehicle identity table corresponding to the vehicle is determined according to the vehicle trajectory, and the vehicle image is obtained at the exit gate as the third vehicle image, from which the identity information is identified. If at least one item is different, a sixth alarm signal is sent to the management end.
2. The intelligent parking lot safety management method according to claim 1, characterized in that: In S1, the method for extracting vehicle features using a vehicle recognition model includes steps S11 to S12. S11, obtaining a vehicle image of the vehicle, and sending it to a license plate recognition model to identify the license plate number. If the license plate number is recognized, the license plate number is used as the vehicle feature of the vehicle. If the license plate number is not recognized, the vehicle image is sent to an unlicensed vehicle recognition model; S12, using the unlicensed vehicle recognition model to recognize the vehicle image, if a small car is recognized from it, using the CILP model to extract the image features of the vehicle image as vehicle features, If a small car is not identified, it will not be allowed to pass. The small car includes two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles with a size smaller than a preset size.
3. The intelligent parking lot safety management method according to claim 2 is characterized by: In S1, the unlicensed vehicle recognition model is trained by the YOLOv8 network based on the small vehicle dataset D1. The samples in the small vehicle dataset D1 are manually labeled vehicle images of two-wheeled vehicles, three-wheeled vehicles, and four-wheeled unlicensed vehicles whose sizes are smaller than a preset size.
4. The intelligent parking lot safety management method according to claim 1, characterized in that: The vehicle collision recognition model is trained by the YOLOv8 network based on the collision data set D2, where the samples in the collision data set D2 are manually labeled human-vehicle collision pictures and vehicle-vehicle collision pictures; The vehicle theft recognition model is obtained by training the YOLOv8 network based on the theft dataset D3, where the samples in the theft dataset D3 are manually marked theft images; The vehicle body color is extracted from the vehicle image by a vehicle body color recognition model.
5. The intelligent parking lot safety management method according to claim 1, characterized in that: S3 also includes, when the management terminal receives the first alarm signal, selecting the vehicle monitoring image and the location of the vehicle determined as a collision event, and sending them to the administrator; S4 also includes, when the management end receives the second alarm signal, obtaining the location of the vehicle and sending it to the administrator.
6. The intelligent parking lot safety management method according to claim 1, characterized in that: S5 also includes, when the management end receives the third alarm signal or the fourth alarm signal, reading the parking space number from the vehicle identity table, obtaining the surveillance videos of 1-3 cameras closest to the parking space number, each video captures a video clip of T2 minutes before the alarm signal, and sending the parking space number and video clip to the administrator.
7. The intelligent parking lot safety management method according to claim 1, characterized in that: S6 also includes, when the management terminal receives the fifth alarm signal or the sixth alarm signal, saving the second vehicle image or the third vehicle image, and querying whether there is vehicle owner information corresponding to the vehicle feature in the management terminal, and notifying the vehicle owner if there is.
