Vehicle safety management method based on intelligent parking lot

Abstract

The present application relates to the technical field of intelligent parking lot, and particularly relates to a vehicle safety management method based on an intelligent parking lot, comprising the following steps: S1, multi-source data acquisition: real-time acquisition of dynamic trajectory data of a target vehicle, parking lot environment parameters and adjacent parking space state data, and generation of a multi-source feature vector set; S2, data space-time alignment: space-time alignment processing of the multi-source feature vector set, and output of a standardized vehicle behavior sequence; S3, abnormal behavior identification: input of the standardized vehicle behavior sequence into an abnormal behavior identification model, generation of an abnormal behavior feature map and a confidence parameter; S4, risk level evaluation: construction of a risk evaluation model based on the confidence parameter and the environment parameter, and output of a risk level under a current scene of the parking lot; and S5, safety strategy triggering: triggering of a corresponding safety strategy execution instruction according to the risk level. The present application can effectively prevent potential risks and ensure the safety of vehicles and personnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent parking lot, and particularly relates to a vehicle safety management method based on intelligent parking lot. BACKGROUND

[0002] With the acceleration of urbanization, parking lots as an important part of urban infrastructure, their management and safety problems are increasingly concerned, especially in large intelligent parking lots, vehicles are dense and the environment is complex, the traditional parking lot safety management method is difficult to cope with the growing safety risks, especially in high-speed and complex urban parking environment, how to monitor the behavior of vehicles in real time and quickly respond to potential safety hazards has become a great challenge, therefore, the parking lot safety management system based on intelligent technology, especially the system that can obtain and analyze vehicle behavior, parking lot environment and parking space state data in real time, has important research significance and application value.

[0003] At present, the existing parking lot safety management method mostly relies on a single monitoring system or a static rule determination mechanism, these methods usually cannot deeply analyze the real-time dynamic behavior of vehicles, cannot identify and handle abnormal behavior in time at the early stage of risk occurrence, for example, the existing system often can only monitor the parking space occupation through static cameras, lacks monitoring and analysis of dynamic information such as vehicle motion trajectory, speed change, acceleration, at the same time, environmental factors such as light, road friction coefficient, etc. are often not considered in real time, leading to risk judgment failure in complex and adverse environments, thereby affecting the efficiency and accuracy of the overall safety management, therefore, the existing technology cannot realize intelligent and dynamic risk assessment and real-time response. SUMMARY

[0004] The present application provides a vehicle safety management method based on intelligent parking lot, which can effectively improve the safety, real-time response ability and management efficiency of the parking lot, reduce potential safety hazards and ensure the safety of personnel, vehicles and facilities.

[0005] The vehicle safety management method based on intelligent parking lot comprises the following steps:

[0006] S1, multi-source data acquisition: real-time acquisition of dynamic trajectory data of target vehicles, parking lot environment parameters and adjacent parking space state data, generation of multi-source feature vector set;

[0007] S2, data space-time alignment: space-time alignment processing of the multi-source feature vector set, output of the standardized vehicle behavior sequence;

[0008] S3, abnormal behavior identification: inputting the standardized vehicle behavior sequence into the abnormal behavior identification model, generating the abnormal behavior feature map and the confidence parameter;

[0009] S4, risk level evaluation: based on the confidence parameter and the environment parameter, a risk evaluation model is constructed, and the risk level of the current scene of the parking lot is output;

[0010] S5, safety policy triggering: according to the risk level, the corresponding safety policy execution instruction is triggered.

[0011] Optionally, the multi-source data collection in S1 includes:

[0012] S11, dynamic trajectory data collection: through TOF ranging of UWB positioning base station and vehicle-mounted label, combined with millimeter wave radar point cloud data, the three-dimensional motion trajectory of the target vehicle is calculated, specifically including:

[0013] The UWB base station array deployed at the top of the column sends ranging signals at a period of 10ms, receives the response signals of the vehicle-mounted label, and calculates the distance d from the base station to the label through the time difference Δt;

[0014] The three-dimensional coordinates (x, y, z) of the vehicle are calculated by using the trilateration algorithm;

[0015] The radial velocity measurement value v of the millimeter wave radar is fused r , and the coordinate jitter is eliminated through Kalman filtering;

[0016] S12, environment parameter acquisition: multi-sensor fusion technology is used to collect the environment characteristics of the parking lot, specifically including:

[0017] Ground friction coefficient calculation: the ground pressure distribution P(x, y) is measured by a piezoelectric sensor array, and the road surface material reflectivity R(λ) is detected by an infrared spectrum analyzer, and the ground friction coefficient μ is calculated;

[0018] Light intensity dynamic compensation: the light intensity of the parking lot is dynamically compensated;

[0019] S13, parking space state monitoring: the vehicle contour recognition algorithm based on YOLOv5 processes multi-view video streams, specifically including:

[0020] Parking space occupation determination: when 5 consecutive frames satisfy , it is determined that the corresponding parking space is occupied, wherein, I edge (x, y) is the edge image gray value of the pixel point (x, y), I total (x, y) is the original image gray value of the pixel point (x, y);

[0021] Safety distance calculation: the safety distance D safe between vehicles is calculated; safe When D<1.5m, an abnormality mark is triggered;

[0022] S14, eigenvector generation: generate a unified format eigenvector through a space-time encoder, specifically including:

[0023] Sliding window alignment: define a sliding window length T=2s, and time synchronize the multi-source data;

[0024] Feature normalization: normalize the multi-source data by using Min-Max normalization processing;

[0025] Eigenvector output: output the multi-source eigenvector set Where Δx, Δy is the instantaneous trajectory change of the vehicle, μ / |v| is the friction to speed ratio, is the light intensity gradient, N danger is satisfied D safe <0.5m of the number of abnormal neighboring vehicles.

[0026] Optionally, the data space-time alignment in S2 includes:

[0027] S21, multi-source timestamp calibration: construct a unified synchronous time axis based on sensor clock bias compensation, specifically including:

[0028] Mark the local timestamps t uwb , t radar , t cam on the UWB positioning data, millimeter wave radar point cloud data, and camera data respectively;

[0029] Calculate the clock bias compensation amount Δt using the PTP protocol;

[0030] Generate a unified reference time axis T sync =t sensor +Δt, where t sensor is the original timestamp of the sensor data;

[0031] S22, space coordinate system unification: establish a global coordinate system of the parking lot and complete the mapping of multi-source data, specifically including:

[0032] Define the global northeast celestial coordinate system origin as O(0,0,0), and set the origin at the geometric center of the parking lot;

[0033] Optimize the UWB base station coordinates by Levenberg-Marquardt;

[0034] Convert the camera data to the global coordinate system through the perspective transformation matrix;

[0035] S23, motion trajectory interpolation reconstruction: use adaptive cubic spline interpolation to compensate for missing data and repair continuity of the trajectory data, and introduce a speed limit condition to optimize the curve shape;

[0036] S24, environment feature correlation modeling: build a vehicle behavior-environment coupling matrix, define an environment influence decay function w(r), and calculate a comprehensive environment parameter E(t);

[0037] S25, standardized sequence generation: perform spatio-temporal dimension normalization processing to generate a standardized vehicle behavior sequence.

[0038] Optionally, the abnormal behavior identification in S3 comprises:

[0039] S31, spatio-temporal feature map generation: using a spatio-temporal graph convolution network (ST-GCN) to extract features from the vehicle behavior sequence, defining key motion nodes (such as car lights, wheels, etc.), combining multi-scale spatio-temporal convolution kernels, capturing vehicle three-dimensional trajectory and time variation, and finally generating a behavior feature map that integrates spatio-temporal characteristics;

[0040] S32, dynamic threshold abnormality determination: introducing an adaptive dynamic threshold mechanism, adjusting the abnormality detection threshold sensitivity according to the environment parameters (such as illumination, friction coefficient), extracting abnormality indicators through max-pooling, and calculating confidence parameters.

[0041] Optionally, the spatio-temporal feature map generation in S31 comprises:

[0042] S311, key motion node definition: defining a vehicle skeleton model to represent key motion nodes of vehicle motion, including chassis motion nodes, body feature nodes, and environment interaction nodes, the chassis motion nodes include left front wheel center N1, right front wheel center N2, left rear wheel center N3, and right rear wheel center N4, the body feature nodes include roof center N5, front and rear bumper midpoints N6, N7, four door hinge points N8, N9, N10, and N11, and the environment interaction nodes include front headlight group centroid N12, tail light group centroid N13, license plate center N14, and charging port position N15, the key motion nodes include three-dimensional coordinates (x, y, z), velocity vector v, and acceleration a. 10 、N 11 , the environment interaction nodes include front headlight group centroid N 12 , tail light group centroid N 13 , license plate center N 14 , charging port position N 15 , the key motion nodes include three-dimensional coordinates (x i ,y i ,z i ), velocity vector v, and acceleration a.

[0043] S312, spatial graph convolution modeling: constructing a spatial correlation matrix between vehicle components and performing graph convolution, defining the connection relationship between vehicle components through an adjacency matrix, considering physical connection and motion similarity, and updating node features in the graph convolution layer to capture the spatial correlation and motion rules between vehicle components, thereby extracting features related to space and time; ​

[0044] S313, multi-scale time convolution: by using dilated convolution, capture behavior features at different time scales;

[0045] S314, spatio-temporal feature fusion: by cross-modal attention mechanism, fuse spatial features and temporal features to form the final spatio-temporal feature atlas.

[0046] Optionally, the dynamic threshold abnormality determination in S32 includes:

[0047] S321, feature channel pooling: multi-dimensional maximum pooling operation is performed on the spatio-temporal feature atlas to extract key abnormal indicators, including acceleration channel c acc and yaw angle channel c yaw ;

[0048] S322, dynamic threshold calculation: construct an environment parameter adaptive dynamic abnormality detection threshold;

[0049] S323, confidence parameter generation: calculate the confidence C based on the relative deviation of the key abnormal indicators and the abnormality detection threshold.

[0050] Optionally, the risk level evaluation in S4 includes:

[0051] S41, risk factor fusion modeling: construct a coupling evaluation model of confidence parameters and environmental parameters, and generate a dynamic risk factor R core ;

[0052] S42, traffic density compensation: introduce real-time traffic density ρ to dynamically adjust the dynamic risk factor;

[0053] S43, time decay correction: apply time decay constraint to sudden risk events to generate the final risk value R final ;

[0054] S44, risk level division: based on the generated final risk value R final , divide the risk level, including low risk, medium risk, high risk, and critical risk.

[0055] Optionally, the risk level division in S44 includes:

[0056] S441, low risk: when R final ≤ 50%, it indicates that the current scene of the parking lot is low risk;

[0057] S442, medium risk: when 50% < R final ≤ 120%, it indicates that the current scene of the parking lot is medium risk;

[0058] S443, high risk: when 120% < R finalWhen the risk is ≤250%, it means the current parking lot scene is medium risk;

[0059] S444, critical risk: When R final When it is >250%, it means that the current parking lot scene is critical risk.

[0060] Optionally, the security policy triggering in S5 includes:

[0061] S51, low risk strategy: When the risk level is low, a voice prompt is sent to the vehicle terminal to remind the driver to pay attention, and the vehicle terminal (OBU) of the target vehicle is used for execution;

[0062] S52, medium risk strategy: When the risk level is medium, activate the sound and light warning devices within a 3-meter radius and enhance the warning effect through intelligent roadside units (RSUs);

[0063] S53, high-risk strategy: When the risk level is high, an obstacle avoidance path is generated and the smart ground lock is controlled to rise through the AR navigation system;

[0064] S54, critical risk strategy: When the risk level is high, the charging piles in the adjacent parking spaces will be powered off, and the physical isolation devices, including hydraulic isolation piers, will be activated for physical safety isolation.

[0065] Beneficial effects of the present invention:

[0066] The present invention, through a vehicle safety management method based on multi-source data collection and intelligent analysis, can obtain the dynamic trajectory data of the target vehicle, parking lot environmental parameters and parking space status data in real time, so as to accurately identify abnormal vehicle behavior and conduct risk assessment in a timely manner. By using spatiotemporal alignment and feature fusion technology, the present invention can ensure a comprehensive analysis of vehicle behavior. Combining the abnormal behavior recognition model with environmental parameters, it can evaluate the risk level of the current parking lot scene in real time.

[0067] The present invention can trigger different security strategies by assessing the risk level to ensure that the risks in the parking lot are effectively controlled. When a low risk is detected, the system will send a voice prompt to the vehicle terminal; when the risk level increases, it will activate the sound and light warning device or control the smart ground lock, and even initiate physical isolation measures in critical situations. Through this intelligent strategy execution mechanism, potential risks can be effectively prevented and the safety of vehicles and personnel can be ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only illustrate the present application, and for those skilled in the art, other drawings can also be obtained without creative effort based on these drawings.

[0069] Fig. 1 The management method flowchart of the embodiment of the present application;

[0070] Fig. 2 The risk level assessment schematic diagram of the embodiment of the present application. DETAILED DESCRIPTION

[0071] The present application will be described in detail below with reference to the drawings and specific embodiments. It should be noted that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be implemented by those skilled in the art for some known technologies; and the drawings are only used to more specifically describe the embodiments, and are not intended to specifically limit the present application.

[0072] It should be noted that in the specification, "one embodiment", "embodiment", "exemplary embodiment", "some embodiments" and the like indicate that the described embodiments can include a specific feature, structure or property, but not necessarily every embodiment includes the specific feature, structure or property. In addition, when a specific feature, structure or property is described in combination with an embodiment, it should be within the knowledge of those skilled in the related art to realize such a feature, structure or property in combination with other embodiments (whether or not explicitly described).

[0073] Generally, the terms can be understood at least in part from the use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular or can be used to describe combinations of features, structures, or characteristics, in the plural. In addition, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but can instead, depending at least in part on the context, allow for the existence of other factors not necessarily explicitly described.

[0074] As Figs. 1-2 The vehicle safety management method based on intelligent parking lot includes the following steps:

[0075] S1, multi-source data acquisition: real-time acquisition of dynamic trajectory data of target vehicle, parking lot environment parameters and adjacent parking space state data, generation of multi-source feature vector set;

[0076] S2, data space-time alignment: space-time alignment processing of multi-source feature vector set, output of standardized vehicle behavior sequence;

[0077] S3, Abnormal behavior recognition: input the standardized vehicle behavior sequence into the abnormal behavior recognition model to generate abnormal behavior feature map and confidence parameter;

[0078] S4, Risk level assessment: based on the confidence parameter and the environment parameter, a risk assessment model is constructed to output the risk level of the current scene in the parking lot;

[0079] S5, Safety policy triggering: according to the risk level, the corresponding safety policy execution instruction is triggered.

[0080] The multi-source data collection in S1 includes:

[0081] S11, Dynamic trajectory data collection: through the TOF ranging of UWB positioning base station and vehicle-mounted label, combined with millimeter wave radar point cloud data, the three-dimensional motion trajectory of the target vehicle is calculated, which specifically includes:

[0082] The UWB base station array deployed at the top of the column sends ranging signals at a period of 10 ms, receives the response signals of the vehicle-mounted label, and calculates the distance d from the base station to the label through the time difference Δt, which is expressed as:

[0083]

[0084] Where c is the speed of light;

[0085] The three-dimensional coordinates (x, y, z) of the vehicle are calculated by using the trilateration algorithm, which is expressed as:

[0086]

[0087] Where (x1, y1, z1), (x2, y2, z2), (x3, y3, z3) are the coordinates of the 1st, 2nd and 3rd base stations respectively, and d1, d2, d3 are the corresponding ranging values;

[0088] The radial velocity measurement value v of the millimeter wave radar is fused r The coordinate jitter is eliminated by Kalman filtering, which is expressed as:

[0089]

[0090] Where, is the vehicle position estimation value after Kalman filtering, α and β are coordinate fusion weight coefficients, is the vehicle position coordinate at the current time calculated by UWB, and θ is the angle between the radar beam and the vehicle motion direction;

[0091] S12, Environment parameter acquisition: multi-sensor fusion technology is used to collect the environment characteristics of the parking lot, which specifically includes:

[0092] Ground friction coefficient calculation: The ground pressure distribution P(x, y) is measured by the piezoelectric sensor array, combined with the infrared spectrum analyzer to detect the road surface material reflectivity R(λ), and the ground friction coefficient μ is calculated, which is expressed as:

[0093]

[0094] Where max(P) is the maximum value in the pressure array, P std is the standard deviation of the pressure data, w1, w2 are weight coefficients in the friction coefficient fusion calculation;

[0095] Light intensity dynamic compensation: The light intensity of the parking lot is dynamically compensated, which is expressed as:

[0096]

[0097] Where L real is the actual light intensity after dynamic compensation, L sensor is the original reading of the light sensor, L dark is the sensor dark current noise, t exposure is the camera exposure time, and K is the exposure compensation calibration coefficient;

[0098] S13, parking space state monitoring: The multi-angle video stream is processed by the vehicle contour recognition algorithm based on YOLOv5, which specifically includes:

[0099] Parking space occupation determination: When the continuous 5 frames satisfy , it is determined that the corresponding parking space is occupied, wherein I edge (x, y) is the edge image gray value of the pixel point (x, y), I total (x, y) is the original image gray value of the pixel point (x, y);

[0100] Safety distance calculation: The safety distance D safe between vehicles is calculated, when D safe <1.5m, the abnormal flag is triggered, and the safety distance D safe is expressed as:

[0101]

[0102] Where (x i , y i ) is the center position coordinate of the i-th vehicle, and (x j , y j ) is the center position coordinate of the j-th vehicle;

[0103] S14, feature vector generation: A unified format feature vector is generated by a space-time encoder, which specifically includes:

[0104] Sliding window alignment: define the sliding window length T = 2s, time synchronization of multi-source data, expressed as:

[0105]

[0106] where t is the current time point, Traj(k) is the trajectory data subset at time k, Env(k) is the environment parameter subset at time k, and Slot(k) is the parking space state parameter subset at time k;

[0107] Feature normalization: Min-Max normalization is used to normalize multi-source data, expressed as:

[0108]

[0109] where, is the normalized i-th feature value, F i is the original feature value, F min , F max are the minimum and maximum values of the original feature value, respectively;

[0110] Feature vector output: output the multi-source feature vector set where Δx, Δy are the instantaneous trajectory changes of the vehicle, μ / |v| is the friction to speed ratio, is the light intensity gradient, N danger satisfies D safe <0.5m of the number of abnormal neighboring vehicles.

[0111] The data space-time alignment in S2 includes:

[0112] S21, multi-source timestamp calibration: construct a unified synchronous time axis based on sensor clock bias compensation, which specifically includes:

[0113] UWB positioning data, millimeter wave radar point cloud data, and camera data are respectively marked with local timestamps t uwb , t radar , t cam ;

[0114] The clock bias compensation amount Δt is calculated using the PTP protocol, expressed as:

[0115]

[0116] where t1, t4 are the master-slave clock synchronization message transmission times, t2, t3 are the slave master clock response times, α ′ is the network delay correction factor, σ delay is the historical delay standard deviation;

[0117] Generate a unified reference time axis Tsync = t sensor + Δt, where t sensor is the original timestamp of the sensor data;

[0118] S22, space coordinate system unification: establish the global coordinate system of the parking lot and complete the multi-source data mapping, which specifically includes:

[0119] Define the origin of the global northeast celestial coordinate system as O(0, 0, 0), and set the origin at the geometric center of the parking lot;

[0120] Optimize the UWB base station coordinates by Levenberg-Marquardt, which is expressed as:

[0121]

[0122] where a i is the initial coordinates of the base station, d i is the measured distance, λ = 1e ；4 is the damping factor, d i is the ranging value, N is the number of base stations in the optimization problem, x is the coordinate point to be optimized, and J is the Jacobian matrix;

[0123] Convert the camera data to the global coordinate system by the perspective transformation matrix, which is expressed as:

[0124]

[0125] where (u, v) is the pixel coordinate, (X, Y) is the mapped ground coordinate, and H is the homography matrix;

[0126] S23, motion trajectory interpolation reconstruction: adopt adaptive cubic spline interpolation to compensate for missing data and repair continuity, and introduce speed limit conditions to optimize the curve shape, which is expressed as:

[0127] S j (t) = a j (t-t j ) 3 +b j (t-t j ) 2 +c j (t-t j )+d j ;

[0128]

[0129]

[0130] where S j (t) is the jth cubic spline interpolation function, aj , b j , c j , d j is the coefficient of the jth interpolation function, t j is the start time of the jth interpolation interval, is the kth derivative of the jth interpolation function at t j:1 , and is the kth derivative of the j+1th interpolation function at t j:1 , and is the acceleration of the interpolation trajectory, a max is the maximum allowed acceleration, dθ / dt is the rate of change of the heading angle, and θ is the vehicle heading angle.

[0131] S24, Environment Feature Association Modeling: Build a vehicle behavior-environment coupling matrix, define the environment influence decay function w(r), and calculate the comprehensive environment parameter E(t), represented as:

[0132]

[0133] where r is the distance from the vehicle to the environment monitoring point, μ(r) is the friction coefficient measured at the point, σ = 3m is the influence radius, μ0 = 0.6 is the reference friction coefficient, L k (t) is the light intensity at the kth point, μ k (t) is the friction coefficient at the kth point, and K is the number of monitoring points involved in the calculation.

[0134] S25, Standardized Sequence Generation: Perform spatiotemporal dimension normalization processing to generate a standardized vehicle behavior sequence, which includes:

[0135] Time downsampling to a uniform frequency of 10Hz, using anti-aliasing sinc interpolation, represented as:

[0136]

[0137] where, is the downsampling value when the sampling frequency is , T is the target uniform sampling period, Δt ′ is the original sampling period, M is the filter order, and x(mΔt ′ ) is the signal value at the mth sampling point in the original data.

[0138] Map the coordinate values to the interval [-1, 1] through spatial normalization, represented as:

[0139]

[0140] where p is the original vehicle three-dimensional coordinate, p min , and p maxs is a historical statistical extreme value range = 2, o bias = -1 is a mapping parameter.

[0141] The abnormal behavior recognition in S3 includes:

[0142] S31, spatiotemporal feature map generation: a spatiotemporal graph convolution network (ST-GCN) is used to extract features from the vehicle behavior sequence, key motion nodes (such as vehicle lights, wheels, etc.) are defined, and multi-scale spatiotemporal convolution kernels are combined to capture the three-dimensional trajectory of the vehicle and the change over time, and finally a behavior feature map that integrates spatiotemporal characteristics is generated;

[0143] S32, dynamic threshold abnormality determination: an adaptive dynamic threshold mechanism is introduced, the abnormality detection threshold sensitivity is adjusted according to environmental parameters (such as illumination, friction coefficient), the abnormality index is extracted through max-pooling, the abnormality detection rate in bad weather is improved, the false alarm rate in sunny high-friction weather is reduced, the abnormality detection rate in rainy low-illumination weather is improved to 89%, the false alarm rate is reduced by 41%, and a confidence parameter is calculated.

[0144] The spatiotemporal feature map generation in S31 includes:

[0145] S311, key motion node definition: a vehicle skeleton model is defined to represent the key motion nodes of vehicle motion, including chassis motion nodes, body feature nodes, and environmental interaction nodes, the chassis motion nodes include the left front wheel center N1, the right front wheel center N2, the left rear wheel center N3, and the right rear wheel center N4, the body feature nodes include the roof center N5, the front and rear bumper midpoints N6 and N7, the four door hinge points N8, N9, and N10, and the environmental interaction nodes include the front headlight group centroid N11, the tail light group centroid N12, the license plate center N13, and the charging port position N14, the key motion nodes include three-dimensional coordinates (x, y, z), velocity vectors 10 , and acceleration 11 12 13 14 15 i i i

[0146] S312, spatial graph convolution modeling: a spatial correlation matrix between vehicle components is constructed and graph convolution is performed, the connection relationship between vehicle components is defined through an adjacency matrix, considering physical connection and motion similarity, the graph convolution layer updates the node features to capture the spatial correlation and motion law between vehicle components, thereby extracting features related to space and time, represented as:

[0147]

[0148] ​​​​​​​​​wherein A ij is an adjacency matrix element, representing the relevance of node i to j, σ ′ is a velocity similarity determination coefficient, is a velocity vector of the node;

[0149]

[0150] Λ ii =∑ j (A+I) ij ;

[0151] wherein H (l:1) is the (l+1)-th graph convolution feature matrix, Λ is the diagonalized degree matrix, A is the adjacency matrix, I is the identity matrix, H (l) is the l-th graph convolution feature matrix, W (l) is the trainable weight matrix of the l-th layer;

[0152] S313, multi-scale time convolution: by using dilated convolution, the behavior features on different time scales are captured, represented as:

[0153]

[0154] wherein T (l:1) (t) is the time convolution output feature of the (l+1)-th layer, the feature value at time step t, W (l) (τ) is the convolution kernel weight of the l-th layer, H (l) (t-d l ·τ) is the input feature matrix of the l-th layer, representing the features at time step t-d l ·τ, k l is the width of the l-th layer convolution kernel, controlling the span of convolution, k l =2 l:1 , d l is the dilation factor of the l-th layer, d l =2 l ;

[0155] S314, spatio-temporal feature fusion: by using cross-modal attention mechanism, the spatial features and the temporal features are fused to form the final spatio-temporal feature atlas, represented as:

[0156]

[0157] wherein M t,i,c is the output after spatio-temporal feature fusion, representing the fusion feature at time step t, spatial node i, feature channel c, α ij is the attention weight, ReLU is the activation function, is the spatial feature and the temporal feature concatenation of the spatial features f is a fusion weight matrix, LeakyReLU is a Leaky ReLU activation function, concatenation of the spatial features and concatenation of the spatial features concatenation of the spatial features and concatenation of the spatial features.

[0158] The dynamic threshold abnormality determination in S32 includes:

[0159] S321, feature channel pooling: multi-dimensional maximum pooling operation is performed on the spatio-temporal feature map to extract key abnormal indicators, including acceleration channel c acc and yaw angle channel c yaw , which is expressed as:

[0160]

[0161] wherein F max is the maximum abnormal indicator after the pooling operation, is the acceleration channel feature value of the t-th time step and the i-th node, is the yaw angle channel feature value of the t-th time step and the i-th node, λ=0.6 is the weight coefficient of the yaw angle, and T is the length of the time window.

[0162] S322, dynamic threshold calculation: a dynamic abnormality detection threshold that is adaptive to environmental parameters is constructed, which is expressed as:

[0163]

[0164] wherein θ dynamic is the abnormality detection threshold, θ base is a sunny standard scene basic threshold, w L is a light intensity influence weight, L curvent is the light intensity of the current scene, L max is the maximum calibration value, indicating the upper limit of the light intensity, w μ is a friction coefficient influence weight, μ base is the reference friction coefficient under dry road surface, μ current is the real-time measured friction coefficient, when L curvent drops or μ current drops, θ dynamic drops to improve the detection sensitivity, when L curvent rises or μ current rises, θ dynamic rises to reduce the false positive rate.

[0165] S323, confidence parameter generation: calculate the confidence C based on the relative deviation of the key abnormal indicators and the abnormal detection threshold, expressed as:

[0166]

[0167] Wherein, C∈[0%,300%] is a hierarchical trigger warning (C>100% is a high-risk event), when μ current <0.3, a 1.2 gain compensation is applied to C.

[0168] The risk level evaluation in S4 includes:

[0169] S41, risk factor fusion modeling: build a coupling evaluation model of confidence parameters and environmental parameters, and generate a dynamic risk factor R core , expressed as:

[0170]

[0171] Wherein, η is the environmental coupling coefficient, k ′ is the attenuation coefficient, R core is the preliminary calculated risk factor, L norm is the normalized light intensity, is the environmental influence gain, is a nonlinear adjustment factor;

[0172] S42, traffic density compensation: introduce real-time traffic density ρ to dynamically adjust the dynamic risk factor, expressed as:

[0173]

[0174] Wherein, R adjusted is the dynamically adjusted risk factor, γ is the weight coefficient of traffic density, ρ base is the reference traffic density, ρ max is the maximum traffic density;

[0175] S43, time decay correction: impose time decay constraint on sudden risk events, generate the final risk value R final , expressed as:

[0176]

[0177] Wherein, R final (t) is the final risk value, λ is the time decay rate, Δt is the time difference between the current time and the starting time of the risk event, is the risk value of the i-th historical risk event, δ is the historical risk decay factor, and s is the number of historical risk events.

[0178] S44, risk level division: based on the generated final risk value R final , the risk level is divided into low risk, medium risk, high risk and critical risk.

[0179] The risk level division in S44 includes:

[0180] S441, low risk: when R final ≤ 50%, indicating that the current scene of the parking lot is low risk;

[0181] S442, medium risk: when 50% < R final ≤ 120%, indicating that the current scene of the parking lot is medium risk;

[0182] S443, high risk: when 120% < R final ≤ 250%, indicating that the current scene of the parking lot is medium risk;

[0183] S444, critical risk: when R final > 250%, indicating that the current scene of the parking lot is critical risk.

[0184] The safety strategy trigger in S5 includes:

[0185] S51, low risk strategy: when the risk level is low risk, a voice prompt is sent to the vehicle terminal to remind the driver to pay attention and execute using the vehicle terminal (OBU) of the target vehicle;

[0186] S52, medium risk strategy: when the risk level is medium risk, the sound and light warning device within 3 meters around is activated, and the warning effect is enhanced through the intelligent roadside unit (RSU);

[0187] S53, high risk strategy: when the risk level is high risk, an obstacle avoidance path is generated, and the intelligent ground lock is controlled to rise through the AR navigation system to ensure the safety of the vehicle;

[0188] S54, critical risk strategy: when the risk level is high risk, the charging pile of the adjacent parking space is powered off, and the physical isolation device including the hydraulic isolation mound is started to perform physical safety isolation to ensure the maximum protection.

[0189] The present application covers any alternative, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0190] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, some improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as falling within the protection scope of the present application.

Claims

1.A method for vehicle safety management based on a smart parking lot, characterized by, The method comprises the following steps: S1, multi-source data collection: real-time acquisition of dynamic trajectory data of the target vehicle, parking lot environment parameters and adjacent parking space state data, generation of a multi-source feature vector set; S2, data space-time alignment: space-time alignment processing of the multi-source feature vector set, output of a standardized vehicle behavior sequence; S3, abnormal behavior identification: input of the standardized vehicle behavior sequence into an abnormal behavior identification model, generation of an abnormal behavior feature map and a confidence parameter; specifically including: S31, space-time feature map generation: feature extraction of the vehicle behavior sequence using a space-time graph convolution network, capture of vehicle three-dimensional trajectory and time variation through definition of key motion nodes and combination of multi-scale space-time convolution kernels, and finally generation of a behavior feature map integrating space-time characteristics; S32, dynamic threshold abnormality determination: introduction of an adaptive dynamic threshold mechanism, adjustment of abnormality detection threshold sensitivity according to environment parameters, extraction of abnormality indicators through maximum pooling, and calculation of a confidence parameter; S4, risk level evaluation: construction of a risk evaluation model based on the confidence parameter and the environment parameter, output of a risk level in the current scene of the parking lot; specifically including: S41, risk factor fusion modeling: a coupling evaluation model of confidence parameters and environmental parameters is constructed, and a dynamic risk factor is generated ; S42, vehicle flow density compensation: introduce real-time vehicle flow density Dynamic risk factor dynamic adjustment is performed; S43, time decay correction: impose time decay constraint on the sudden risk event to generate the final risk value ; S44, risk level classification: based on the generated final risk value , the risk level classification includes low risk, medium risk, high risk, and critical risk; S5, safety strategy triggering: triggering of corresponding safety strategy execution instructions according to the risk level. 2.The smart parking lot based vehicle safety management method of claim 1, wherein, The multi-source data collection in S1 comprises: S11, dynamic trajectory data collection: calculation of the three-dimensional motion trajectory of the target vehicle through TOF ranging of the UWB positioning base station and the vehicle-mounted label, and combination of millimeter wave radar point cloud data; specifically including: The UWB base station array deployed at the top of the column sends ranging signals at a 10ms cycle, receives the response signals of the vehicle-mounted tags, and calculates the distance from the base station to the tags through the time difference Calculate the distance from the base station to the tag ; Adopting three-side positioning algorithm to solve three-dimensional coordinates of vehicle ; Fusing radial velocity measurements of millimeter wave radar by kalman filtering S12, environment parameter acquisition: acquisition of parking lot environment characteristics using multi-sensor fusion technology; specifically including: Ground friction coefficient calculation: ground pressure distribution is measured by piezoelectric sensor array , combined with infrared spectrum analyzer to detect road surface material reflectivity , calculate ground friction coefficient ; Light intensity dynamic compensation: dynamic compensation of the light intensity of the parking lot; S13, parking space state monitoring: processing of multi-view video streams based on a YOLOv5 vehicle contour recognition algorithm; specifically including: Parking space occupation determination: when 5 consecutive frames satisfy , it is determined that the corresponding parking space is occupied, wherein, is the edge image gray value of the pixel point , is the original image gray value of the pixel point ; Safety gap calculation: calculating a safety gap between vehicles When an exception flag is triggered; S14, feature vector generation: generation of a uniform format feature vector through a space-time encoder; specifically including: Sliding window alignment: defining sliding window length Time synchronizing the multi-source data; Feature normalization: normalization processing of multi-source data using Min-Max normalization processing; Feature vector output: output a set of multi-source feature vectors wherein, is a change in instantaneous trajectory of the vehicle, is a friction to speed ratio, is a lighting intensity gradient, is a number of abnormal neighboring vehicles satisfying . 3.The smart parking lot based vehicle safety management method of claim 1, wherein, The data space-time alignment in S2 comprises: S21, multi-source timestamp calibration: construction of a unified synchronous time axis based on sensor clock bias compensation; specifically including: The UWB positioning data, millimeter wave radar point cloud data, and camera data are respectively marked with local time stamps , , ; Calculating clock deviation compensation amount using ptp protocol ; Generating a unified reference timeline wherein, is a raw timestamp of the sensor data; S22, unification of space coordinate systems: establishment of a parking lot global coordinate system and completion of multi-source data mapping; specifically including: The origin of the global northeast sky coordinate system is defined as and is set at the parking lot geometric center; Levenberg-Marquardt optimization of UWB base station coordinates; Conversion of camera data to the global coordinate system through a perspective transformation matrix; S23, motion trajectory interpolation reconstruction: missing compensation and continuity repair of trajectory data using adaptive cubic spline interpolation, and introduction of a speed limit condition to optimize the curve shape; S24, environment feature correlation modeling: build a vehicle behavior-environment coupling matrix, define environment influence decay function , and calculate the comprehensive environment parameter ; S25, standardized sequence generation: execution of space-time dimension normalization processing, generation of a standardized vehicle behavior sequence. 4.The smart parking lot based vehicle safety management method of claim 1, wherein, The space-time feature map generation in S31 comprises: S311, key motion node definition: defining the key motion nodes of the vehicle skeleton model representing the vehicle motion, including chassis motion nodes, body feature nodes, environment interaction nodes, the chassis motion nodes including the center of the left front wheel , the center of the right front wheel , the center of the left rear wheel , the center of the right rear wheel , the body feature nodes including the center of the roof , the midpoint of the front and rear bumpers , , the four-door hinge points , , , , the environment interaction nodes including the center of mass of the front headlight group , the center of mass of the tail light group , the center of the license plate , the charging port position , the key motion nodes including three-dimensional coordinates , velocity vectors , acceleration ; S312, spatial graph convolution modeling: construct the spatial correlation matrix between vehicle components and perform graph convolution, define the connection relationship between vehicle components through adjacency matrix, consider physical connection and motion similarity, update node features through graph convolution layer to capture spatial correlation and motion law between vehicle components, and extract features related to space and time; S313, multi-scale time convolution: capture behavior features at different time scales by using dilated convolution; S314, spatio-temporal feature fusion: fuse spatial features and temporal features through cross-modal attention mechanism to form the final spatio-temporal feature map. 5.The smart parking lot based vehicle safety management method of claim 4, wherein, The dynamic threshold abnormality determination in S32 includes: S321, feature channel pooling: multi-dimensional maximum pooling operation is performed on the spatio-temporal feature map to extract key abnormal indicators, including acceleration channel with yaw angle channel ; S322, dynamic threshold calculation: construct a dynamic determination abnormality detection threshold adaptive to environmental parameters; S323, confidence parameter generation: calculate the confidence based on the relative deviation of the key abnormal indicator and the abnormal detection threshold . 6.The smart parking lot based vehicle safety management method of claim 1, wherein, The risk level division in S44 includes: S441, low risk: when the current scene of the parking lot is low risk; S442, medium risk: when the current scene of the parking lot is medium risk; S443, high risk: when the current scene of the parking lot is high risk; S444, critical risk: when the current scene of the parking lot is a critical risk. 7.The smart parking lot based vehicle safety management method of claim 6, wherein, The safety policy triggering in S5 includes: S51, low-risk strategy: when the risk level is low, send a voice prompt to the vehicle terminal to remind the driver to pay attention and execute using the vehicle terminal of the target vehicle; S52, medium-risk strategy: when the risk level is medium, activate the sound and light warning device within 3 meters around, and enhance the warning effect through the intelligent roadside unit; S53, high-risk strategy: when the risk level is high, generate an obstacle avoidance path and control the intelligent ground lock to rise through the AR navigation system; S54, critical risk strategy: when the risk level is high, disconnect the charging pile of the adjacent parking space and start the physical isolation device, including the hydraulic isolation mound for physical safety isolation.

Images