Parking lot parking space monitoring system and method

Abstract

The invention relates to the technical field of parking lot parking space monitoring, in particular to a parking lot parking space monitoring system and method. The battery car is identified through a battery car standard parking detection model, a battery car standard parking coefficient is obtained, and the standard degree of battery car parking is judged; identifying a fire feature of the battery car through a battery car anomaly detection model to obtain a battery car anomaly coefficient; the fire coefficient of the battery car is obtained by combining the environment temperature, the battery service life ratio and the continuous charging duration; the identity type of the battery car interaction personnel is judged through face recognition, and an interaction personnel identity abnormity coefficient is obtained; performing action recognition through an interaction personnel behavior anomaly detection model to obtain an interaction personnel behavior anomaly coefficient; and combining the time point and the duration of the interaction to obtain a comprehensive abnormal coefficient of the interaction personnel. Effective and safe management of the storage battery car parking lot is realized through the storage battery car standard parking coefficient, the fire coefficient of the storage battery car and the comprehensive abnormal coefficient.

Description

Technical Field

[0001] This invention relates to the field of parking space monitoring technology, specifically a parking space monitoring system and method. Background Technology

[0002] As a flexible and economical mode of transportation, electric bicycles are gradually being accepted by the public and have become one of the main choices for short-distance travel. With the increasing number of electric bicycles owned by residents, residential communities are building well-equipped parking lots and installing charging stations to facilitate the use and parking of these bicycles.

[0003] At the same time, the increase in the number of electric bicycles has also led to increasingly serious safety issues. Improper parking, fires, and battery theft are common occurrences, seriously endangering the lives and property of residents and affecting the safety and standardized management of electric bicycle parking lots.

[0004] Existing security technologies for electric bicycle parking lots rely primarily on cameras for monitoring, which is inefficient and inaccurate. These cameras often serve only as evidence after an accident occurs and fail to effectively manage the safety of electric bicycle parking lots.

[0005] Based on the above factors, this invention proposes a parking space monitoring system and method. Summary of the Invention

[0006] The purpose of this invention is to provide a parking lot monitoring system and method. The system identifies electric bicycles using a standardized parking detection model to obtain a standardized parking coefficient, thus determining the degree of standardized parking. It also identifies fire characteristics of electric bicycles using an anomaly detection model to obtain an anomaly coefficient. Combined with ambient temperature, battery lifespan, and continuous charging time, a fire coefficient is obtained. Furthermore, it determines the identity type of the person interacting with the electric bicycle using facial recognition, obtaining an anomaly coefficient for the person's identity. Finally, it performs action recognition using an anomaly detection model for the person's behavior, obtaining an anomaly coefficient for the person's behavior. Finally, it combines the time and duration of the interaction to obtain a comprehensive anomaly coefficient for the person interacting with the electric bicycle. This achieves effective and safe management of electric bicycle parking lots.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A parking lot space monitoring system, characterized in that it includes an electric bicycle data acquisition module, an electric bicycle parking monitoring module, an electric bicycle fire monitoring module, and an electric bicycle anti-theft monitoring module;

[0009] The electric vehicle data acquisition module is used to acquire electric vehicle data information and electric vehicle owner data information;

[0010] The electric bicycle parking monitoring module measures the degree of proper parking of electric bicycles through an electric bicycle proper parking coefficient; at the same time, a first threshold is set, and when the electric bicycle proper parking coefficient is greater than the first threshold, a message is sent to the electric bicycle owner and the community property management; the electric bicycle proper parking coefficient is determined by an electric bicycle proper parking detection model;

[0011] The electric vehicle fire monitoring module is used to calculate the electric vehicle fire coefficient. The electric vehicle fire coefficient is calculated by the electric vehicle anomaly coefficient, battery life ratio, battery continuous charging time and ambient temperature. A second threshold is set. If the electric vehicle fire coefficient exceeds the second threshold, an early warning is issued. The electric vehicle anomaly coefficient is obtained by detecting and recognizing the infrared detection image of the electric vehicle through an electric vehicle anomaly detection model.

[0012] The electric bicycle anti-theft monitoring module identifies and judges the personnel interacting with the electric bicycle; when the personnel are the electric bicycle owner or their family members, the process is terminated; when the personnel are not the electric bicycle owner or their family members, a comprehensive abnormality coefficient is calculated based on the abnormality coefficient of the personnel's identity, the abnormality coefficient of the personnel's behavior, the abnormality coefficient of the interaction time point, and the interaction time length; this coefficient is then compared with a set third threshold to determine whether an abnormality exists.

[0013] The electric vehicle data acquisition module acquires the following data information: electric vehicle color, model, battery model, and battery life.

[0014] The electric vehicle data acquisition module acquires the following electric vehicle owner data information: owner image information, owner contact information, owner account number, and owner family member image information.

[0015] The electric vehicle proper parking detection model detects and judges the proper parking of the electric vehicle;

[0016] The electric bicycle standardized parking detection model includes an electric bicycle parking image acquisition layer, an electric bicycle parking image preprocessing layer, an electric bicycle parking image feature extraction layer, an electric bicycle parking distance recognition layer, an electric bicycle parking angle recognition layer, an electric bicycle tilt degree recognition layer, an electric bicycle carried item recognition layer, a feature fusion processing layer, and a recognition result output layer.

[0017] The electric bicycle parking image acquisition layer uses multiple cameras installed in the community's electric bicycle parking lot to identify and photograph the electric bicycles, thereby obtaining a set of parking detection images of the electric bicycles.

[0018] The electric vehicle parking image preprocessing layer is used to preprocess the parking detection image set to obtain a preprocessed parking detection image set;

[0019] The electric vehicle parking image feature extraction layer is used to extract image features from the preprocessed parking detection image set to obtain an image feature dataset.

[0020] The electric vehicle parking spacing recognition layer is used to identify the parking spacing of the electric vehicles through the image feature dataset to obtain a parking spacing coefficient; the electric vehicle parking spacing includes the spacing between the electric vehicle and each border of the designated parking space, the spacing between the electric vehicles, and the spacing between the electric vehicle and the charging pile.

[0021] The electric vehicle parking angle recognition layer identifies the parking angle of the electric vehicle through the image feature dataset to obtain the parking angle coefficient;

[0022] The electric vehicle tilt recognition layer identifies the tilt degree of the electric vehicle through the image feature dataset and obtains the tilt degree coefficient;

[0023] The electric vehicle carrying item recognition layer identifies the items carried by the electric vehicle through the image feature dataset to obtain the electric vehicle carrying coefficient;

[0024] The feature fusion processing layer obtains the standard parking coefficient of the electric vehicle through the parking spacing coefficient, the parking angle coefficient, the tilt coefficient, and the electric vehicle load coefficient;

[0025] The recognition result output layer is used to output the standard parking coefficient of the electric vehicle.

[0026] The steps for obtaining the abnormality coefficient of the electric vehicle are as follows:

[0027] The electric bicycles are photographed in real time by an infrared thermal imager installed in the electric bicycle parking lot to obtain infrared detection images of the electric bicycles.

[0028] The infrared detection image of the electric vehicle is input into the electric vehicle anomaly detection model for identification, and the anomaly coefficient of the electric vehicle is obtained.

[0029] The electric vehicle anomaly detection model identifies changes in the temperature and smoke of the electric vehicle.

[0030] The electric vehicle anomaly detection model includes an infrared image input layer, an infrared image preprocessing layer, an infrared image feature extraction layer, a first factor identification module, a second factor identification module, and an anomaly coefficient output layer.

[0031] The infrared image input layer is used to input the infrared detection image into the model for training;

[0032] The infrared image preprocessing layer is used to preprocess the infrared detection image to obtain a preprocessed infrared detection image;

[0033] The infrared image feature extraction layer is used to extract relevant data features of the electric vehicle from the preprocessed infrared detection image;

[0034] The first factor recognition module is used to identify the temperature of the electric vehicle in the infrared detection image and obtain the first factor training result;

[0035] The second factor recognition module is used to identify the smoke emitted by the electric vehicle based on the changes in the continuous infrared detection images, and to obtain the second factor training result;

[0036] The anomaly coefficient output layer obtains the anomaly coefficient of the electric vehicle based on the training results of the first factor and the training results of the second factor, and outputs the anomaly coefficient.

[0037] The electric vehicle fire coefficient is used to measure the probability of the electric vehicle catching fire;

[0038] The formula for calculating the fire ignition coefficient of the electric vehicle is as follows:

[0039]

[0040] Wherein; FC represents the electric vehicle fire coefficient, and the larger the value of FC, the greater the probability of the electric vehicle catching fire; AC represents the electric vehicle abnormality coefficient; SL represents the battery life ratio; CT represents the continuous charging time of the battery; AT represents the ambient temperature of the electric vehicle; δ is the set temperature threshold; and exp represents an exponential function with base e.

[0041] The calculation process for the anomaly coefficient of the interactive personnel's identity and the anomaly coefficient of the interactive personnel's behavior is as follows:

[0042] The identity information of the interacting person is identified through the facial image of the interacting person;

[0043] The detection process is terminated when the person interacting is the owner of the electric vehicle or their family member.

[0044] When the person interacting is not the owner of the electric vehicle or his / her family, further identification and detection are performed on the person to obtain the abnormality coefficient of the person interacting;

[0045] The abnormal behavior of the interacting personnel is then identified using an interactive personnel behavior anomaly detection model to obtain the abnormal behavior coefficient of the interacting personnel.

[0046] The electric bicycle anti-theft monitoring module uses the comprehensive anomaly coefficient to judge and warn of electric bicycle theft.

[0047] The comprehensive anomaly coefficient is calculated based on the relevant data information of the people interacting with the electric vehicle; it includes the anomaly coefficient of the person's identity, the anomaly coefficient of the person's behavior, the anomaly coefficient of the interaction time point, and the interaction time length.

[0048] The formula for calculating the comprehensive anomaly coefficient is as follows:

[0049]

[0050] Wherein; SA represents the comprehensive anomaly coefficient of the interacting person; AI represents the anomaly coefficient of the interacting person's identity; AB represents the anomaly coefficient of the interacting person's behavior; TP represents the anomaly coefficient of the time point at which the interacting person interacts with the electric vehicle, and the value of TP varies depending on the time classification; TL represents the interaction time length of the interacting person interacting with the electric vehicle; ε represents the set interaction time threshold; and exp represents an exponential function with base e.

[0051] A parking lot space monitoring method, characterized in that:

[0052] Obtain data on electric bicycles and their owners within the residential community;

[0053] Multiple cameras installed in the community's electric bicycle parking lot are used to identify and photograph the electric bicycles, resulting in a set of parking detection images of the electric bicycles. The set of parking detection images includes images of the parking spacing of the electric bicycles, images of the parking angle of the electric bicycles, images of the tilt degree of the electric bicycles, and images of the items carried by the electric bicycles.

[0054] The set of images of the electric bicycles being parked is input into the electric bicycle standardized parking detection model to identify and determine the standardized parking coefficient of the electric bicycles; the standardized parking coefficient of the electric bicycles is compared with the set first threshold, and a message is sent to the electric bicycle owners and the community property management.

[0055] The electric bicycle is detected using an infrared thermal imager to obtain an infrared detection image; the infrared detection image is input into an electric bicycle anomaly detection model for training to obtain an electric bicycle anomaly coefficient; the electric bicycle fire coefficient is calculated based on the electric bicycle anomaly coefficient, battery life ratio, battery continuous charging time, and ambient temperature; a second threshold is set, and if the electric bicycle fire coefficient exceeds the second threshold, a warning is simultaneously issued to the community property management and the vehicle owner;

[0056] The facial recognition model is used to verify the identity of the person interacting with the electric vehicle, and the abnormality coefficient of the person's identity is obtained. The abnormality detection model of the person's behavior is used to detect the behavior of the person interacting with the vehicle, and the abnormality coefficient of the person's behavior is obtained. The comprehensive abnormality coefficient is calculated based on the abnormality coefficient of the person's identity, the abnormality coefficient of the person's behavior, the abnormality coefficient of the interaction time point, the interaction time length, and the set interaction duration threshold, to determine whether theft has occurred and to issue an early warning.

[0057] The electric vehicle fire coefficient is used to measure the probability of the electric vehicle catching fire;

[0058] The formula for calculating the fire ignition coefficient of the electric vehicle is as follows:

[0059]

[0060] Wherein, FC represents the electric vehicle fire coefficient, and the larger the value of FC, the greater the probability of the electric vehicle catching fire; AC represents the electric vehicle anomaly coefficient, which is obtained by the electric vehicle anomaly detection model through the identification of temperature and smoke changes of the electric vehicle; SL represents the battery life ratio; CT represents the continuous charging time of the battery; AT represents the ambient temperature of the electric vehicle; δ is the set temperature threshold; and exp represents an exponential function with base e.

[0061] The probability of electric bicycle theft is measured by a comprehensive anomaly coefficient.

[0062] The comprehensive anomaly coefficient is calculated based on relevant data information of the personnel interacting with the electric vehicle; it includes anomaly coefficients for the identity of the personnel, anomaly coefficients for the behavior of the personnel, anomaly coefficients for the interaction time point, and interaction time length; and it is used to determine whether there is any electric vehicle theft.

[0063] The formula for calculating the comprehensive anomaly coefficient is as follows:

[0064]

[0065] Wherein; SA represents the comprehensive anomaly coefficient of the interacting personnel; AI represents the anomaly coefficient of the interacting personnel's identity; AB represents the anomaly coefficient of the interacting personnel's behavior; TP represents the anomaly coefficient of the interaction time point between the interacting personnel and the electric vehicle, and the value of TP varies depending on the time classification; TL represents the interaction time length between the interacting personnel and the electric vehicle; ε represents the set interaction time threshold; and exp represents an exponential function with base e.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] 1. This invention uses a standardized parking detection model for electric bicycles to identify four aspects: parking distance, parking angle, tilt degree, and cargo load. Four training results are obtained: parking distance coefficient, parking angle coefficient, tilt degree coefficient, and cargo load coefficient. These training results are then used to derive a standardized parking coefficient for the electric bicycles. This invention uses a standardized parking detection model to perform multi-angle detection and identification of electric bicycles, obtaining a standardized parking coefficient that accurately measures the degree of standardized parking and strengthens the management of standardized parking in electric bicycle parking lots.

[0068] 2. This invention identifies the main characteristics of electric bicycle fires using an anomaly detection model, obtaining an anomaly coefficient to effectively judge and quantify changes in electric bicycle temperature and smoke. Combined with the ambient temperature, battery lifespan, and continuous charging time, the fire coefficient of the electric bicycle is calculated. The main characteristics of electric bicycle fires include changes in electric bicycle temperature and smoke. This invention calculates the fire coefficient by considering the fire coefficient, ambient temperature, battery lifespan, and continuous charging time, accurately measuring the probability of electric bicycle fires and strengthening fire prevention and early warning management in electric bicycle parking lots.

[0069] 3. The system uses facial recognition to determine the identity type of the person interacting with the electric bicycles, assigning coefficients based on the identity type to obtain an anomaly coefficient for the person's identity. It then uses an anomaly detection model to identify the actions of the person interacting with the bicycles, obtaining an anomaly coefficient for their behavior. Finally, it combines the interaction time and duration to calculate a comprehensive anomaly coefficient for the person interacting with the bicycles. This invention achieves accurate identification of theft behavior by people interacting with the bicycles, strengthening the anti-theft early warning management of electric bicycle parking lots. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0071] Figure 2 This is a schematic diagram of the electric vehicle standardized parking detection model structure of the present invention;

[0072] Figure 3 This is a schematic diagram of the electric vehicle anomaly detection model structure of the present invention;

[0073] Figure 4 This is a schematic diagram illustrating the violent combustion of the electric vehicle according to the present invention;

[0074] Figure 5 This is a schematic diagram illustrating the large amount of smoke generated by the electric vehicle of the present invention;

[0075] Figure 6 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] With the increasing number of electric bicycles owned by residents in residential communities, the issues of parking and charging within these communities urgently need to be addressed. Many residential communities have built well-equipped parking lots and installed charging stations to facilitate charging for electric bicycle owners. However, due to the lack of effective monitoring technology in these parking lots, risks such as improper parking, fires, and battery theft still exist, seriously affecting the safety of electric bicycle owners and residents.

[0078] Based on the above factors, this invention proposes a parking space monitoring system and method.

[0079] The system and method monitor parking spaces in a parking lot using cameras and infrared thermal imaging instruments.

[0080] This invention uses Hikvision's 5A dome-shaped star network camera, specifically model DS-2CD5A25EWDV2-IZ(S)(JM). This device features a built-in motorized zoom lens, is easy to operate, and provides smooth zooming. It boasts a maximum resolution of 2 megapixels and can output real-time images at 30fps. It also features fog penetration and electronic image stabilization, making it ideal for capturing and detecting electric bicycles in parking lots.

[0081] This invention employs a Hikvision H7 dual-spectrum tube camera, specifically model DS-2TD2667-35 / P, for temperature and smoke detection in electric bicycles. The infrared device has a thermal imaging pixel size of 640×512, a pixel size of 17μm, and a net density (NETD) of <35mK (25℃, F1.0). It supports functions such as area intrusion detection, boundary crossing detection, area entry detection, area exit detection, audio anomaly detection, and high-temperature object detection. Temperature measurement, fire detection, and smoke detection functions can be activated simultaneously. It exhibits good sensitivity and accuracy in identifying fire characteristics of electric bicycles.

[0082] The following detailed implementation examples illustrate the solution of this invention.

[0083] Example 1

[0084] Residential Community A is an older community completed 20 years ago, and the parking and charging of electric bicycles were not considered in the initial design. In recent years, with the continuous increase in the number of electric bicycles owned by residents and tenants, the problem of charging and parking has gradually become prominent. Many residents, for convenience, choose to park their electric bicycles in stairwells or fire lanes; they connect charging strips to the bicycles through windows, doors, etc. This charging method poses a significant safety hazard, and there have been numerous cases of electric bicycles spontaneously combusting in stairwells, elevators, and other residential areas across the country. To eliminate the safety hazards of electric bicycles and strengthen the safety and standardized management of electric bicycles within the community, under the consultation of the community committee and property management, Residential Community A plans to build multiple electric bicycle parking lots within the community for residents' electric bicycle charging and parking; at the same time, it will adopt the parking space monitoring system described in this invention for safety monitoring.

[0085] The structure of the parking space monitoring system is as follows: Figure 1 As shown, it includes an electric bicycle data acquisition module, an electric bicycle parking monitoring module, an electric bicycle fire monitoring module, and an electric bicycle anti-theft monitoring module;

[0086] The electric vehicle data acquisition module is used to acquire electric vehicle data information and electric vehicle owner data information within residential community A.

[0087] The electric vehicle data acquisition module acquires electric vehicle data information including: electric vehicle color, model, battery model, and battery life; the electric vehicle owner data acquired by the electric vehicle data acquisition module includes electric vehicle owner data: owner's family members, owner's contact information, owner's account number, and owner's image.

[0088] The data on electric bicycles and their owners in residential community A will be stored in a database for comparison and identification during subsequent monitoring.

[0089] To accurately and effectively manage the parking lot for electric bicycles in the community, this invention collects information from both the electric bicycles and their owners. By collecting relevant information about the electric bicycles and understanding their battery usage, it can help in the accurate calculation of the electric bicycle fire risk. By collecting information about the electric bicycle owners, it can help in the identity verification of the anti-theft monitoring module and better calculate the comprehensive anomaly coefficient of the interacting personnel.

[0090] The electric bicycle parking monitoring module measures the degree of proper parking of electric bicycles in residential community A by using an electric bicycle proper parking coefficient; at the same time, a first threshold is set, and when the electric bicycle proper parking coefficient is greater than the first threshold, a message is sent to the electric bicycle owner and the community property management.

[0091] After the electric bicycle parking lot in residential community A was completed and put into use, in order to ensure that residents park their electric bicycles in accordance with regulations, this invention proposes an electric bicycle standardized parking coefficient; the electric bicycle standardized parking coefficient is obtained by image recognition detection through an electric bicycle standardized parking detection model.

[0092] The electric bicycle standardized parking detection model is obtained by structural improvement and adjustment based on the YOLOv7 model; the structure of the electric bicycle standardized parking detection model is as follows: Figure 2 As shown.

[0093] The electric bicycle standardized parking detection model includes an electric bicycle parking image acquisition layer, an electric bicycle parking image preprocessing layer, an electric bicycle parking image feature extraction layer, an electric bicycle parking distance recognition layer, an electric bicycle parking angle recognition layer, an electric bicycle tilt degree recognition layer, an electric bicycle carried item recognition layer, a feature fusion processing layer, and a recognition result output layer.

[0094] The electric bicycle parking image acquisition layer uses multiple cameras installed in the community's electric bicycle parking lot to identify and photograph the electric bicycles, thereby obtaining a set of parking detection images of the electric bicycles.

[0095] The electric vehicle parking image preprocessing layer is used to preprocess the parking detection image set to obtain a preprocessed parking detection image set;

[0096] The electric vehicle parking image feature extraction layer is used to extract image features from the preprocessed parking detection image set to obtain an image feature dataset.

[0097] The electric vehicle parking spacing recognition layer is used to identify the parking spacing of the electric vehicle through the image feature dataset to obtain a parking spacing coefficient; the electric vehicle parking spacing includes the spacing between the electric vehicle and each border of the designated parking space, the spacing between the electric vehicle and other electric vehicles, and the spacing between the electric vehicle and the charging pile.

[0098] The electric vehicle parking angle recognition layer identifies the parking angle of the electric vehicle through the image feature dataset to obtain the parking angle coefficient;

[0099] The electric vehicle tilt recognition layer identifies the tilt degree of the electric vehicle through the image feature dataset and obtains the tilt degree coefficient;

[0100] The electric vehicle carrying item recognition layer identifies the items carried by the electric vehicle through the image feature dataset to obtain the electric vehicle carrying coefficient;

[0101] The feature fusion processing layer obtains the standard parking coefficient of the electric vehicle through the parking spacing coefficient, the parking angle coefficient, the tilt coefficient, and the electric vehicle load coefficient;

[0102] The recognition result output layer is used to output the standard parking coefficient of the electric vehicle.

[0103] This invention uses monitoring cameras to capture and identify electric bicycles in a parking lot, obtaining a set of parking detection images that include four aspects: parking distance, parking angle, tilt degree, and cargo load. Then, the electric bicycle standardized parking detection model is used to detect and identify the parking detection image set to obtain the electric bicycle standardized parking coefficient, which can accurately measure the degree of standardized parking of the electric bicycles.

[0104] The electric vehicle fire monitoring module is used to calculate the fire coefficient of electric vehicles in the parking lot inside residential community A; and to measure whether the electric vehicle is in a fire state; the electric vehicle fire coefficient is calculated by the electric vehicle anomaly coefficient, battery age ratio, battery continuous charging time and ambient temperature; at the same time, a second threshold is set, and if the electric vehicle fire coefficient exceeds the second threshold, an alert is issued to the community property management and the vehicle owner; the electric vehicle anomaly coefficient is obtained by detecting and recognizing the infrared detection image of the electric vehicle through the electric vehicle anomaly detection model.

[0105] Among the safety hazards of electric bicycles, battery fires are the most dangerous. This invention analyzes and summarizes multiple cases of electric bicycle battery fires, extracting several factors closely related to these fires: the battery's lifespan, continuous charging time, and ambient temperature. Furthermore, this invention uses infrared thermal imaging equipment to monitor and capture images of electric bicycles in parking lots, obtaining infrared detection images. Then, an anomaly detection model is used to identify and train features such as temperature and smoke during an electric bicycle fire, resulting in an anomaly coefficient.

[0106] The steps for obtaining the abnormality coefficient of the electric vehicle are as follows:

[0107] The electric vehicles are photographed in real time by an infrared thermal imager set up in the electric vehicle parking lot to obtain infrared detection images of the electric vehicles.

[0108] The infrared detection image of the electric vehicle is input into the electric vehicle anomaly detection model for identification, and the anomaly coefficient of the electric vehicle is obtained.

[0109] The model structure of the electric vehicle anomaly detection model is as follows: Figure 3 As shown.

[0110] The electric vehicle anomaly detection model is obtained by structural improvement and adjustment based on the classic convolutional neural network (CNN); it includes an infrared image input layer, an infrared image preprocessing layer, an infrared image feature extraction layer, a first factor recognition module, a second factor recognition module, and an anomaly coefficient output layer.

[0111] The infrared image input layer is used to input the infrared detection image into the model for training;

[0112] The infrared image preprocessing layer is used to preprocess the infrared detection image to obtain a preprocessed infrared detection image;

[0113] The infrared image feature extraction layer is used to extract relevant data features of the electric vehicle from the preprocessed infrared detection image;

[0114] The first factor recognition module is used to identify the temperature of the electric vehicle in the infrared detection image to obtain the first factor training result; the violent burning phenomenon of the electric vehicle is as follows: Figure 4 As shown;

[0115] The second factor recognition module is used to identify the smoke emitted by the electric vehicle by continuously detecting changes in the infrared images, and to obtain the second factor training result; the phenomenon of the electric vehicle producing a large amount of smoke is as follows: Figure 5 As shown;

[0116] The anomaly coefficient output layer obtains the anomaly coefficient of the electric vehicle based on the training results of the first factor and the training results of the second factor, and outputs the anomaly coefficient.

[0117] This invention identifies the main characteristics of electric vehicle fires using an anomaly detection model, and obtains an anomaly coefficient for the electric vehicle. The main characteristics of electric vehicle fires include changes in electric vehicle temperature and changes in electric vehicle smoke; thus, it achieves effective judgment and quantification of changes in electric vehicle temperature and smoke.

[0118] After obtaining the electric vehicle anomaly coefficient, the electric vehicle fire coefficient is further calculated using the electric vehicle anomaly coefficient, battery life ratio, battery continuous charging time, and ambient temperature.

[0119] The electric vehicle fire coefficient is used to measure the likelihood of the electric vehicle catching fire.

[0120] The formula for calculating the fire ignition coefficient of the electric vehicle is as follows:

[0121]

[0122] Wherein; FC represents the electric vehicle fire coefficient, and the larger the value of FC, the greater the probability of the electric vehicle catching fire; AC represents the electric vehicle abnormality coefficient; SL represents the battery service life ratio, which is the ratio of the actual service life to the set service life; CT represents the continuous charging time of the battery; AT represents the ambient temperature of the electric vehicle; δ is the set temperature threshold; and exp represents an exponential function with base e.

[0123] When determining the probability of the electric vehicle catching fire, the present invention obtains the electric vehicle's anomaly coefficient by acquiring the electric vehicle's temperature and smoke data through an infrared thermal imaging device; and then, by combining the ambient temperature of the electric vehicle at that time, the battery's service life ratio, and the battery's continuous charging time, accurately calculates the electric vehicle's fire coefficient.

[0124] The following calculation of the fire coefficient of electric vehicles is based on the data in Table 1; the temperature threshold is set to 25℃.

[0125] Table 1. Relevant Data on Fire Initiation Factors in Residential Community A's Electric Vehicle Parking Lot

[0126] electric bicycle Electric vehicle abnormality coefficient Battery lifespan ratio Continuous charging time Ambient temperature Temperature threshold A1 0.23 1.18 3.00 hours 30℃ 25℃ A2 0.18 0.22 4.80 hours 30℃ 25℃ A3 0.34 0.54 2.45 hours 30℃ 25℃

[0127] The formula and result for calculating the fire ignition factor of electric bicycle A1 are as follows:

[0128]

[0129] The formula and result for calculating the fire ignition factor of electric bicycle A3 are as follows:

[0130]

[0131] Based on the above calculations, in the electric bicycle parking lot of residential community A, the fire ignition coefficient of electric bicycle A1 is 3.0511; and the fire ignition coefficient of electric bicycle A3 is 2.0812.

[0132] In addition to regulating the parking and fire detection of electric bicycles, this invention also considers the risk of theft. It categorizes electric bicycle theft into three types: theft of goods carried on the bicycle, theft of bicycle parts, and the theft of the entire bicycle. Among these, due to the small size and ease of disassembly of electric bicycle batteries, the rapid sale of used electric bicycle batteries, and their relatively high resale value, electric bicycle batteries are a more suitable target for theft compared to the larger and somewhat bulkier electric bicycles themselves.

[0133] The electric bicycle anti-theft monitoring module identifies and judges the personnel interacting with the electric bicycle; when the personnel interacting are the electric bicycle owner and their family members, the process is terminated; when the personnel interacting are not the electric bicycle owner and their family members, the module calculates the comprehensive anomaly coefficient to determine whether theft has occurred and issues an early warning.

[0134] The comprehensive anomaly coefficient is calculated based on relevant data information of the personnel interacting with the electric bicycle; it includes the anomaly coefficient of the personnel's identity, the anomaly coefficient of the personnel's behavior, the time point of the interaction, and the duration of the interaction; and it is used to determine whether there is any electric bicycle theft.

[0135] The formula for calculating the comprehensive anomaly coefficient is as follows:

[0136]

[0137] Wherein, SA represents the comprehensive anomaly coefficient of the interacting person; AI represents the anomaly coefficient of the interacting person's identity; AB represents the anomaly coefficient of the interacting person's behavior; TP represents the anomaly coefficient of the time point at which the interacting person interacts with the electric vehicle, and the value of TP varies depending on the time classification; TL represents the interaction time length of the interacting person interacting with the electric vehicle; ε represents the set interaction time threshold; and exp represents an exponential function with base e.

[0138] This invention verifies the identity of people interacting with electric bicycles to obtain an anomaly coefficient; it monitors the behavior of people interacting with electric bicycles using an anomaly detection model to identify suspicious actions and obtain a real-time anomaly coefficient; and it calculates the comprehensive anomaly coefficient of people interacting with electric bicycles by combining the time point and duration of the interaction. This allows for accurate identification, classification, and early warning of people interacting with electric bicycles.

[0139] The calculation process for the abnormality coefficient of the interactive personnel's identity and the abnormality coefficient of the interactive personnel's behavior is as follows:

[0140] The identity information of the interacting person is identified through the facial image of the interacting person;

[0141] The detection process is terminated when the person interacting is the owner of the electric vehicle or their family member.

[0142] When the person interacting is not the owner of the electric vehicle or his / her family, further identification and detection are performed on the person to obtain the abnormality coefficient of the person interacting;

[0143] The abnormal behavior of the interacting personnel is identified by an interactive personnel behavior anomaly detection model, and the abnormal behavior coefficient of the interacting personnel is obtained.

[0144] The interactive human behavior anomaly detection model and facial recognition model are obtained by structural improvements and adjustments based on the classic convolutional neural network (CNN).

[0145] Facial recognition is used to determine the type of person interacting, including car owners, car owners' family members, community residents, people who appear in person, and people who do not appear in person, and different coefficients are assigned to them.

[0146] The identification process terminates when the person interacting is identified as the owner of an electric bicycle or a family member of the owner.

[0147] When it is determined that the person interacting is not the owner of the electric bicycle or a family member of the owner, the person interacting is further divided into community residents, people who appear in public, and people who do not appear in public.

[0148] The residents of the community are those who are not owners of electric bicycles and their families, including other residents and the community property management staff.

[0149] The individuals who appeared were non-residents of the community who had previously been present in the community and whose images were recorded in the community's database; these included delivery drivers, couriers, real estate agents, and maintenance workers.

[0150] The person who did not appear in person was neither a resident of the community nor had their image information left in the community's database.

[0151] This invention verifies the identity of personnel interacting with electric bicycles and obtains an anomaly coefficient. Based on the identity verification result, it further identifies the actions of the interacting personnel through an anomaly detection model and obtains an anomaly coefficient. Through identity verification and behavior verification, it effectively classifies and judges the personnel interacting with electric bicycles.

[0152] The following is a display of the comprehensive anomaly coefficients using specific scenario data. Specifically, in assigning values ​​to the identity types of individuals, the anomaly coefficient for community residents is set to 0.3; the anomaly coefficient for individuals who have appeared in public is set to 0.6; and the anomaly coefficient for individuals who have not appeared in public is set to 0.9. In assigning values ​​to the anomaly coefficients for interaction times, the coefficients for 8:00 to 18:00 are set to 0.2; the coefficients for 18:00 to 0:00 are set to 0.4; and the coefficients for 0:00 to 8:00 are set to 0.8. The threshold for interaction duration is set to 6 minutes.

[0153] Table 2. Data related to the comprehensive anomaly coefficient of the electric bicycle parking lot in Residential Community A.

[0154] Interactive personnel Identity Anomaly Coefficient Behavioral abnormality coefficient Interactive Time Points Interaction duration A1 A11 0.3 0.26 8：02 1 minute A2 A21 0.6 0.52 19：52 5 minutes A3 A31 0.3 0.18 12：09 4 minutes A3 A32 0.9 0.36 4：30 4 minutes

[0155] The formula and result for calculating the comprehensive anomaly coefficient of interactive personnel A21 are as follows:

[0156]

[0157] The formula and result for calculating the comprehensive anomaly coefficient of interactive personnel A32 are as follows:

[0158]

[0159] Based on the above calculations, the overall anomaly coefficient of interactive user A21 is 0.3311, and the overall anomaly coefficient of interactive user A32 is 0.4109.

[0160] After using the parking space monitoring system proposed in this invention for a period of time, residential community A achieved good results. A point system was established for the proper parking of electric bicycles, and these points can be used for free charging of the bicycles. When the parking space monitoring system detects improper parking of an electric bicycle, it sends a message to the community property management and the owner. If the owner has not parked the vehicle properly after a preset time, points are deducted according to the degree of improper parking and the duration of the improper parking. Significant results have also been achieved in detecting electric bicycle fires and preventing theft, preventing two cases of electric bicycle battery spontaneous combustion and one case of electric bicycle battery theft, issuing timely alarms, reducing residents' property losses, and protecting their lives.

[0161] This invention uses a battery-powered vehicle standardized parking detection model to identify four aspects: parking distance, parking angle, tilt degree, and cargo load. This yields four training results: parking distance coefficient, parking angle coefficient, tilt degree coefficient, and cargo load coefficient. These training results are then used to derive a standardized parking coefficient, accurately measuring the degree of standardized parking. An abnormality detection model identifies the main characteristics of battery-powered vehicle fires, obtaining an abnormality coefficient. This enables effective judgment and quantification of temperature and smoke changes. Combined with the ambient temperature, battery life, and continuous charging time, a fire coefficient is calculated, accurately measuring the probability of a battery-powered vehicle fire. The system uses facial recognition to determine the identity type of the person interacting with the electric vehicle, and assigns a coefficient based on the identity type to obtain the anomaly coefficient of the person's identity. The system also uses an anomaly detection model to identify the actions of the person interacting with the vehicle, and obtains the anomaly coefficient of the person's behavior. Finally, the system combines the time point and duration of the interaction to calculate the comprehensive anomaly coefficient of the person interacting with the vehicle, thereby achieving accurate judgment of theft behavior by the person interacting with the vehicle.

[0162] Effective and safe management of electric vehicle parking lots is achieved by using electric vehicle standardized parking coefficients, electric vehicle fire coefficients, and comprehensive anomaly coefficients.

[0163] Example 2

[0164] Residential Community B is a newly completed and handed-over community in recent years. Its pricing and apartment layouts cater to low- to middle-income buyers. In addition to parking for motor vehicles, it also includes above-ground and underground parking facilities for electric bicycles. However, in practice, many residents, for convenience, park their electric bicycles haphazardly at the entrance of the parking lots, resulting in low space utilization and posing risks to property management and security. Furthermore, the risks of fires from prolonged charging and battery theft remain, seriously threatening the lives and property of residents in Residential Community B.

[0165] In order to better manage the non-motorized vehicle parking lot in the community, residential community B uses the parking space monitoring method described in this invention to improve residents' awareness of proper parking of electric bicycles and strengthen fire prevention and theft prevention supervision.

[0166] The specific steps of the parking space monitoring method are as follows: Figure 6 As shown.

[0167] Obtain data on electric bicycles and their owners within residential community B;

[0168] The electric vehicle data acquisition module acquires data including: electric vehicle color, model, battery model, and battery life; and electric vehicle owner data, including: owner's family members, owner's contact information, owner's household registration number, and owner's image. The acquired data on electric vehicles and their owners in residential community B is stored in a database for future comparison and identification during monitoring.

[0169] Multiple cameras installed in the non-motorized vehicle parking lot B of the residential community are used to identify and photograph the electric bicycles, resulting in a set of parking detection images of the electric bicycles. The set of parking detection images includes images of the parking distance of the electric bicycles, images of the parking angle of the electric bicycles, images of the tilt degree of the electric bicycles, and images of the items carried by the electric bicycles.

[0170] The set of images of the electric bicycles being parked is input into the electric bicycle standardized parking detection model to identify and determine the standardized parking coefficient of the electric bicycles; the standardized parking coefficient of the electric bicycles is compared with the set first threshold, and a message is sent to the electric bicycle owners and the community property management.

[0171] The electric bicycle standardized parking detection model includes an electric bicycle parking image acquisition layer, an electric bicycle parking image preprocessing layer, an electric bicycle parking image feature extraction layer, an electric bicycle parking distance recognition layer, an electric bicycle parking angle recognition layer, an electric bicycle tilt degree recognition layer, an electric bicycle carried item recognition layer, a feature fusion processing layer, and a recognition result output layer.

[0172] By using surveillance cameras to capture and identify electric bicycles in the parking lot, a set of parking detection images is obtained, including four aspects: parking distance, parking angle, tilt degree, and cargo status of electric bicycles. Then, the electric bicycle standardized parking detection model is used to detect and identify the parking detection image set to obtain the electric bicycle standardized parking coefficient, which can accurately measure the degree of standardized parking of the electric bicycles.

[0173] The electric vehicle is detected using an infrared thermal imager to obtain an infrared detection image; the infrared detection image is input into the electric vehicle anomaly detection model for training to obtain the electric vehicle anomaly coefficient; the electric vehicle fire coefficient is calculated based on the electric vehicle anomaly coefficient, battery life ratio, battery continuous charging time and ambient temperature; a second threshold is set, and if the electric vehicle fire coefficient exceeds the second threshold, a warning is simultaneously issued to the community property management and the vehicle owner.

[0174] The electric vehicle fire coefficient is used to measure the likelihood of the electric vehicle catching fire.

[0175] The formula for calculating the fire ignition coefficient of the electric vehicle is as follows:

[0176]

[0177] Wherein; FC represents the electric vehicle fire coefficient, and the larger the value of FC, the greater the probability of the electric vehicle catching fire; AC represents the electric vehicle abnormality coefficient; SL represents the battery life ratio; CT represents the continuous charging time of the battery; AT represents the ambient temperature of the electric vehicle; δ is the set temperature threshold; and exp represents an exponential function with base e.

[0178] The following calculation of the fire coefficient of the electric vehicle is based on the data in Table 3; the temperature threshold is set to 25℃.

[0179] Table 3. Relevant Data on Fire Incidence Coefficient of Electric Bicycle Parking Lot in Residential Community B

[0180] electric bicycle Electric vehicle abnormality coefficient Battery lifespan ratio Continuous charging time Ambient temperature Temperature threshold B1 0.16 0.23 5.00 hours 32℃ 25℃ B2 0.25 0.49 3.60 hours 32℃ 25℃ B3 0.27 0.92 2.28 hours 32℃ 25℃

[0181] The formula and result for calculating the fire ignition factor of electric vehicle B1 are as follows:

[0182]

[0183] The formula and result for calculating the fire ignition factor of electric vehicle B3 are as follows:

[0184]

[0185] Based on the above calculations, in the non-motorized vehicle parking lot of residential community B, the fire ignition coefficient of electric bicycle B1 is 1.9445; and the fire ignition coefficient of electric bicycle B3 is 2.6725.

[0186] The facial recognition model is used to verify the identity of the person interacting with the electric vehicle, and the abnormality coefficient of the person's identity is obtained; the abnormal behavior detection model is used to detect the behavior of the person interacting with the vehicle, and the abnormal behavior coefficient of the person interacting with the vehicle is obtained.

[0187] The comprehensive anomaly coefficient is calculated based on the anomaly coefficients of the interacting person's identity, behavior, time of interaction, duration of interaction, and set interaction duration threshold to determine whether theft has occurred and to issue an early warning.

[0188] The formula for calculating the comprehensive anomaly coefficient is as follows:

[0189]

[0190] Wherein; SA represents the comprehensive anomaly coefficient of the interacting person; AI represents the anomaly coefficient of the interacting person's identity; AB represents the anomaly coefficient of the interacting person's behavior; TP represents the anomaly coefficient of the time point at which the interacting person interacts with the electric vehicle, and the value of TP varies depending on the time classification; TL represents the interaction time length of the interacting person interacting with the electric vehicle; ε represents the set interaction time threshold; and exp represents an exponential function with base e.

[0191] The calculation process for the abnormality coefficient of the interactive personnel's identity and the abnormality coefficient of the interactive personnel's behavior is as follows:

[0192] The identity information of the interacting person is identified through the facial image of the interacting person;

[0193] The detection process is terminated when the person interacting is the owner of the electric vehicle or their family member.

[0194] When the person interacting is not the owner of the electric vehicle or his / her family, further identification and detection are performed on the person to obtain the abnormality coefficient of the person interacting;

[0195] The abnormal behavior of the interacting personnel is identified by an interactive personnel behavior anomaly detection model, and the abnormal behavior coefficient of the interacting personnel is obtained.

[0196] Facial recognition is used to determine the type of person interacting, including car owners, car owners' family members, community residents, people who appear in person, and people who do not appear in person, and different coefficients are assigned to them.

[0197] The identification process terminates when the person interacting is identified as the owner of an electric bicycle or a family member of the owner.

[0198] When it is determined that the person interacting is not the owner of the electric bicycle or a family member of the owner, the person interacting is further divided into community residents, people who appear in public, and people who do not appear in public.

[0199] The residents of the community are those who are not owners of electric bicycles and their families, including other residents and the community property management.

[0200] The individuals who appeared were non-residents of the community who had previously been present in the community and whose images were recorded in the community's database; these included delivery drivers, couriers, real estate agents, and maintenance workers.

[0201] The person who did not appear in person was neither a resident of the community nor had their image information left in the community's database.

[0202] The actions of the interacting personnel are identified by the abnormal behavior detection model, and the abnormal behavior coefficient of the interacting personnel is obtained. The interacting personnel of the electric vehicle are effectively classified and judged through identity verification and behavior verification.

[0203] The following section presents a comprehensive anomaly coefficient display using specific scenario data;

[0204] Specifically, in assigning values ​​to the types of personnel identities, the identity anomaly coefficient for community residents is set to 0.3; the identity anomaly coefficient for people who appear in public is set to 0.6; and the identity anomaly coefficient for people who do not appear in public is set to 0.9. In assigning values ​​to the interaction time points, the time from 8:00 to 18:00 is set to 0.2; the time from 18:00 to 0:00 is set to 0.4; and the time from 0:00 to 8:00 is set to 0.8. The interaction duration threshold is set to 5 minutes.

[0205] Table 4. Data related to the comprehensive anomaly coefficient of the electric bicycle parking lot in Residential Community B.

[0206] Interactive personnel Identity Anomaly Coefficient Behavioral abnormality coefficient Interactive Time Points Interaction duration B1 B11 0.6 0.14 20：22 4 minutes B1 B12 0.6 0.32 16：20 3 minutes B2 B21 0.3 0.25 9：30 5 minutes B3 B31 0.3 0.44 2：45 6 minutes B3 B32 0.6 0.23 6:50 2 minutes

[0207] The formula and result for calculating the comprehensive anomaly coefficient of interactive personnel B12 are as follows:

[0208]

[0209] The formula and result for calculating the comprehensive anomaly coefficient of interactive personnel B32 are as follows:

[0210]

[0211] Based on the above calculations, the overall anomaly coefficient for interactive user B12 is 0.1189, and the overall anomaly coefficient for interactive user B32 is 0.1035.

[0212] This invention identifies electric bicycles using a standardized parking detection model to obtain a standardized parking coefficient; it also determines the degree of standardized parking; it identifies fire characteristics of electric bicycles using an anomaly detection model to obtain an anomaly coefficient; and it combines this with ambient temperature, battery life ratio, and continuous charging time to obtain a fire coefficient; it uses facial recognition to determine the identity type of people interacting with the electric bicycles to obtain an anomaly coefficient; it uses an anomaly detection model to perform action recognition to obtain an anomaly coefficient; and it combines the time and duration of the interaction to obtain a comprehensive anomaly coefficient; thus achieving effective and safe management of electric bicycle parking lots.

[0213] The embodiments of the present invention have been shown and described above. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments without departing from the principles and spirit of the present invention are all within the protection scope of the present invention, which is defined by the appended claims and their equivalents.

Claims

1. A parking lot space monitoring system, characterized in that: It includes a battery vehicle data acquisition module, a battery vehicle parking monitoring module, a battery vehicle fire monitoring module, and a battery vehicle anti-theft monitoring module; The electric vehicle data acquisition module is used to acquire electric vehicle data information and electric vehicle owner data information; The electric bicycle parking monitoring module measures the degree of proper parking of electric bicycles through an electric bicycle proper parking coefficient; at the same time, a first threshold is set, and when the electric bicycle proper parking coefficient is greater than the first threshold, a message is sent to the electric bicycle owner and the community property management; the electric bicycle proper parking coefficient is determined by an electric bicycle proper parking detection model; The electric vehicle fire monitoring module is used to calculate the electric vehicle fire coefficient. The electric vehicle fire coefficient is calculated by the electric vehicle anomaly coefficient, battery life ratio, continuous battery charging time, and ambient temperature. At the same time, a second threshold is set. If the electric vehicle fire coefficient exceeds the second threshold, an early warning is issued. The electric vehicle anomaly coefficient is obtained by detecting and recognizing the infrared detection image of the electric vehicle through an electric vehicle anomaly detection model. The electric bicycle anti-theft monitoring module identifies and judges the personnel interacting with the electric bicycle; when the personnel are the electric bicycle owner or their family members, the process is terminated; when the personnel are not the electric bicycle owner or their family members, a comprehensive abnormality coefficient is calculated based on the abnormality coefficient of the personnel's identity, the abnormality coefficient of the personnel's behavior, the abnormality coefficient of the interaction time point, and the interaction time length; this coefficient is then compared with a set third threshold to determine whether an abnormality exists.

2. The parking space monitoring system according to claim 1, characterized in that: The electric vehicle data acquisition module acquires the following data information: electric vehicle color, model, battery model, and battery life. The electric vehicle data acquisition module acquires the following electric vehicle owner data information: owner image information, owner contact information, owner account number, and owner family member image information.

3. The parking space monitoring system according to claim 1, characterized in that: The electric vehicle proper parking detection model detects and judges the proper parking of the electric vehicle; The electric bicycle standardized parking detection model includes an electric bicycle parking image acquisition layer, an electric bicycle parking image preprocessing layer, an electric bicycle parking image feature extraction layer, an electric bicycle parking distance recognition layer, an electric bicycle parking angle recognition layer, an electric bicycle tilt degree recognition layer, an electric bicycle carried item recognition layer, a feature fusion processing layer, and a recognition result output layer. The electric bicycle parking image acquisition layer uses multiple cameras installed in the community's electric bicycle parking lot to identify and photograph the electric bicycles, thereby obtaining a set of parking detection images of the electric bicycles. The electric vehicle parking image preprocessing layer is used to preprocess the parking detection image set to obtain a preprocessed parking detection image set; The electric vehicle parking image feature extraction layer is used to extract image features from the preprocessed parking detection image set to obtain an image feature dataset. The electric vehicle parking spacing recognition layer is used to identify the parking spacing of the electric vehicles through the image feature dataset to obtain a parking spacing coefficient; the electric vehicle parking spacing includes the spacing between the electric vehicle and each border of the designated parking space, the spacing between the electric vehicles, and the spacing between the electric vehicle and the charging pile. The electric vehicle parking angle recognition layer identifies the parking angle of the electric vehicle through the image feature dataset to obtain the parking angle coefficient; The electric vehicle tilt recognition layer identifies the tilt degree of the electric vehicle through the image feature dataset and obtains the tilt degree coefficient; The electric vehicle carrying item recognition layer identifies the items carried by the electric vehicle through the image feature dataset to obtain the electric vehicle carrying coefficient; The feature fusion processing layer obtains the standard parking coefficient of the electric vehicle through the parking spacing coefficient, the parking angle coefficient, the tilt coefficient, and the electric vehicle load coefficient; The recognition result output layer is used to output the standard parking coefficient of the electric vehicle.

4. The parking space monitoring system according to claim 1, characterized in that: The steps for obtaining the abnormality coefficient of the electric vehicle are as follows: The electric bicycles are photographed by an infrared thermal imager installed in the electric bicycle parking lot to obtain infrared detection images of the electric bicycles. The infrared detection image of the electric vehicle is input into the electric vehicle anomaly detection model for identification, and the anomaly coefficient of the electric vehicle is obtained. The electric vehicle anomaly detection model identifies changes in the temperature and smoke of the electric vehicle. The electric vehicle anomaly detection model includes an infrared image input layer, an infrared image preprocessing layer, an infrared image feature extraction layer, a first factor identification module, a second factor identification module, and an anomaly coefficient output layer. The infrared image input layer is used to input the infrared detection image into the model for training; The infrared image preprocessing layer is used to preprocess the infrared detection image to obtain a preprocessed infrared detection image; The infrared image feature extraction layer is used to extract relevant data features of the electric vehicle from the preprocessed infrared detection image; The first factor recognition module is used to identify the temperature of the electric vehicle in the infrared detection image and obtain the first factor training result; The second factor recognition module is used to identify the smoke emitted by the electric vehicle based on the changes in the continuous infrared detection images, and to obtain the second factor training result; The anomaly coefficient output layer obtains the anomaly coefficient of the electric vehicle based on the training results of the first factor and the training results of the second factor, and outputs the anomaly coefficient.

5. A parking space monitoring system according to claim 1, characterized in that: The electric vehicle fire coefficient is used to measure the probability of the electric vehicle catching fire; The formula for calculating the fire ignition coefficient of the electric vehicle is as follows: in; FC represents the electric vehicle fire coefficient. The larger the value of FC, the greater the probability of the electric vehicle catching fire. AC represents the electric vehicle abnormality coefficient. SL represents the battery service life ratio, which is the ratio of the actual service life of the battery to the set service life. CT represents the continuous charging time of the battery. AT represents the ambient temperature of the electric vehicle. δ is the set temperature threshold. exp represents an exponential function with base e.

6. A parking space monitoring system according to claim 1, characterized in that: The calculation process for the anomaly coefficient of the interactive personnel's identity and the anomaly coefficient of the interactive personnel's behavior is as follows: The identity information of the interacting person is identified through the facial image of the interacting person; The detection process is terminated when the person interacting is the owner of the electric vehicle or their family member. When the person interacting is not the owner of the electric vehicle or his / her family, further identification and detection are performed on the person to obtain the abnormality coefficient of the person interacting; The abnormal behavior of the interacting personnel is then identified using an interactive personnel behavior anomaly detection model to obtain the abnormal behavior coefficient of the interacting personnel.

7. A parking space monitoring system according to claim 1, characterized in that: The electric bicycle anti-theft monitoring module uses the comprehensive anomaly coefficient to judge and warn of electric bicycle theft. The comprehensive anomaly coefficient is calculated based on the relevant data information of the people interacting with the electric vehicle; it includes the anomaly coefficient of the person's identity, the anomaly coefficient of the person's behavior, the anomaly coefficient of the interaction time point, and the interaction time length. The formula for calculating the comprehensive anomaly coefficient is as follows: in; SA represents the overall anomaly coefficient of the interacting person; AI represents the anomaly coefficient of the interacting person's identity; AB represents the anomaly coefficient of the interacting person's behavior; TP represents the anomaly coefficient at the time point when the interacting person interacts with the electric vehicle, and the value of TP varies depending on the interaction time; TL represents the interaction time length of the interacting person interacting with the electric vehicle; ε represents the set interaction time threshold; exp represents an exponential function with base e.

8. A parking lot space monitoring method, characterized in that: Obtain data on electric bicycles and their owners within the residential community; Multiple cameras installed in the community's electric bicycle parking lot are used to identify and photograph the electric bicycles, resulting in a set of parking detection images of the electric bicycles. The set of parking detection images includes images of the parking spacing of the electric bicycles, images of the parking angle of the electric bicycles, images of the tilt degree of the electric bicycles, and images of the items carried by the electric bicycles. The set of images of the electric bicycles being parked is input into the electric bicycle standardized parking detection model to identify and determine the standardized parking coefficient of the electric bicycles; the standardized parking coefficient of the electric bicycles is compared with the set first threshold, and a message is sent to the electric bicycle owners and the community property management. The electric vehicle is detected by an infrared thermal imager to obtain an infrared detection image; the infrared detection image is input into the electric vehicle anomaly detection model for training to obtain the electric vehicle anomaly coefficient; the electric vehicle fire coefficient is calculated based on the electric vehicle anomaly coefficient, battery life ratio, battery continuous charging time and ambient temperature. A second threshold is set. If the fire coefficient of the electric vehicle exceeds the second threshold, a warning will be issued to both the community property management and the vehicle owner. The facial recognition model is used to verify the identity of the people interacting with the electric vehicle, and the anomaly coefficient of the person interacting with the vehicle is obtained. The behavior of the interacting personnel is detected by an abnormal behavior detection model to obtain the abnormal behavior coefficient of the interacting personnel; then, the comprehensive abnormal coefficient is calculated based on the abnormal coefficient of the interacting personnel's identity, the abnormal coefficient of the interacting personnel's behavior, the abnormal coefficient of the interaction time point, the interaction time length, and the set interaction duration threshold to determine whether there is abnormal behavior.

9. A parking space monitoring method according to claim 8, characterized in that: The electric vehicle fire coefficient is used to measure the probability of the electric vehicle catching fire; The formula for calculating the fire ignition coefficient of the electric vehicle is as follows: in; FC represents the electric vehicle fire coefficient. The larger the FC value, the greater the probability of the electric vehicle catching fire. AC represents the electric vehicle anomaly coefficient, which is obtained by the electric vehicle anomaly detection model through the identification of temperature and smoke changes of the electric vehicle. SL represents the battery service life ratio, which is the ratio of the actual service life of the battery to the set service life. CT represents the continuous charging time of the battery. AT represents the ambient temperature of the electric vehicle. δ is the set temperature threshold. exp represents an exponential function with base e.

10. A parking space monitoring method according to claim 8, characterized in that: The probability of electric bicycle theft is measured by a comprehensive anomaly coefficient. The comprehensive anomaly coefficient is calculated based on relevant data information of the personnel interacting with the electric vehicle; it includes the anomaly coefficient of the personnel's identity, the anomaly coefficient of the personnel's behavior, the anomaly coefficient of the interaction time point, and the interaction time length; and it determines whether there is any abnormal behavior. The formula for calculating the comprehensive anomaly coefficient is as follows: in; SA represents the overall anomaly coefficient of the interacting personnel; AI represents the anomaly coefficient of the interacting personnel's identity; AB represents the anomaly coefficient of the interacting personnel's behavior; TP represents the anomaly coefficient of the interaction time point when the interacting personnel interacts with the electric vehicle, and the value of TP varies depending on the time classification; TL represents the interaction time length when the interacting personnel interacts with the electric vehicle; ε represents the set interaction time threshold; exp represents an exponential function with base e.

Images