Underground parking lot management system based on multi-sensor fusion and computer equipment

Abstract

The invention discloses an underground parking lot management system based on multi-sensor fusion and computer equipment, and the system comprises a parking space recommendation and navigation module which is used for receiving a parking request of a user, and generating a parking space recommendation result and an initial navigation path through a weighted scoring algorithm based on a preset evaluation index; the data acquisition module is used for continuously acquiring multi-source detection data including vehicle positioning data, parking space state sensing data and visual identification data after the vehicle enters the monitoring area of the parking lot; the fusion judgment module is used for fusing the multi-source data by adopting a preset fusion algorithm and judging the state of the recommended parking space in real time; and the dynamic scheduling module is used for responding to the condition that the parking space state becomes unavailable, re-executing the weighted scoring algorithm to generate a new recommended parking space and updating a navigation path. The technical problem that dynamic guide adjustment cannot be realized when the parking space state is abnormal due to inaccurate perception and decision lag in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent parking management, in particular to an underground parking lot management system based on multi-sensor fusion and a computer device. BACKGROUND

[0002] The current underground parking lot management system generally detects the parking space state through a single sensor such as an ultrasonic wave or a camera, and provides initial parking space guidance for the vehicle owner based on this static information. However, such a system has obvious limitations: due to the lack of real-time dynamic tracking and comprehensive judgment of the parking lot after the vehicle enters, the system cannot effectively deal with sudden conditions such as temporary occupation of the initial recommended parking space, vehicle following entry, or human occupation, resulting in guidance failure and increasing the time and energy consumption of the vehicle owner to find a parking space. In addition, the sensors of the existing system usually work independently, and it is difficult to reliably determine the real state of the parking space in complex environments, and it is easy to cause false judgments due to shielding, light changes or sensor false alarms. The above problems make it difficult for the existing parking lot management system to realize true intelligence and real-time, therefore, there is an urgent need for an underground parking lot management system that can fuse multi-source sensor data, accurately perceive the state change of the parking space in real time, and dynamically plan the path. SUMMARY

[0003] The main purpose of the present application is to provide an underground parking lot management system based on multi-sensor fusion and a computer device, which aims to solve the technical problem that the existing technology cannot realize dynamic guidance adjustment when the parking space state is abnormal due to inaccurate perception and decision lag.

[0004] To achieve the above purpose, the first aspect of the present application provides an underground parking lot management system based on multi-sensor fusion, comprising: a parking space recommendation and navigation module, configured to receive a user parking request, generate a parking space recommendation result through a weighted scoring algorithm based on a preset evaluation index, and generate an initial navigation path from the current location of the user to the recommended parking space; a data acquisition module, configured to continuously acquire multi-source detection data from multiple types of sensors in response to detecting that the user's vehicle enters a preset monitoring area of the parking lot, wherein the multi-source detection data includes vehicle positioning data, parking space state perception data, and visual recognition data; a fusion determination module, configured to fuse the multi-source detection data according to a preset parking space data fusion algorithm, and determine the state of the recommended parking space in real time; a dynamic scheduling module, configured to respond to the state of the recommended parking space becoming unavailable, re-execute the weighted scoring algorithm based on real-time multi-source detection data and a preset evaluation index, generate a new parking space recommendation result, and update the navigation path.

[0005] Preferably, the data acquisition module comprises: a data triggering unit configured to trigger a data collection process when detecting that a user vehicle enters the preset monitoring area; a positioning data acquisition unit configured to acquire vehicle positioning data reported by a user terminal; the user terminal calculates the vehicle positioning data based on a positioning algorithm by receiving signals of a plurality of Bluetooth beacons in a parking lot; a radar data collection unit configured to collect parking space state perception data by a radar sensor arranged above a parking space, the parking space state perception data including existence state, motion state and contour feature information of a target object on the parking space; a visual data collection unit configured to collect visual data by a camera module arranged in the parking lot, and generate visual recognition data including parking space number and license plate number recognition results by performing image recognition processing on the visual data using a target detection model based on a convolutional neural network.

[0006] Preferably, the positioning data acquisition unit comprises: a signal receiving subunit configured to control the user terminal to receive signal strength data from a plurality of Bluetooth beacons; a filtering processing subunit configured to process the signal strength data using a Kalman filtering algorithm to filter out noise and improve positioning stability; a positioning solution subunit configured to calculate vehicle position coordinates based on the processed signal strength data by a positioning solution model to generate the vehicle positioning data.

[0007] Preferably, the fusion determination module comprises: a probability assignment generation unit configured to generate a plurality of basic probability assignments corresponding to parking space occupancy states based on the vehicle positioning data, the parking space state perception data and the visual recognition data; an evidence fusion unit configured to perform fusion calculation on the plurality of basic probability assignments based on a Dempster combination rule to obtain an integrated trust degree that the recommended parking space is occupied; a state determination unit configured to compare the integrated trust degree with a preset determination threshold, and determine the state of the recommended parking space according to a comparison result.

[0008] Preferably, the probability assignment generation unit comprises: a first assignment subunit configured to calculate a distance relationship between a user vehicle and the recommended parking space based on the vehicle positioning data to generate a first basic probability assignment; a second assignment subunit configured to generate a second basic probability assignment based on target existence state and motion features in the parking space state perception data; A third allocation subunit is configured to generate a third basic probability allocation based on the parking space number and license plate number recognition result in the visual recognition data.

[0009] Preferably, the evidence fusion unit comprises: A conflict calculation subunit is configured to calculate a conflict coefficient between the first basic probability allocation, the second basic probability allocation, and the third basic probability allocation. A synthesized evidence generation subunit is configured to synthesize all the evidences supporting the "recommended parking space is occupied" state in the plurality of basic probability allocations to obtain a preliminary synthesized evidence, and calculate a normalization factor using the conflict coefficient. A comprehensive probability allocation subunit is configured to divide the preliminary synthesized evidence by the normalization factor to obtain a fused comprehensive basic probability allocation. A trust degree extraction subunit is configured to extract a probability value corresponding to the "recommended parking space is occupied" state from the comprehensive basic probability allocation as the comprehensive trust degree.

[0010] Preferably, the parking space recommendation and navigation module comprises: A request receiving unit is configured to receive a parking request submitted by a user through a mobile terminal, wherein the request comprises destination information and parking space preference type. A parking space screening unit is configured to obtain real-time parking space state data of a parking lot and screen out a set of available parking spaces. A score calculation unit is configured to perform weighted score calculation on each parking space in the set of available parking spaces based on distance from the destination, path planning distance, estimated time consumption, and parking space preference matching degree. A recommendation generation unit is configured to sort the parking spaces according to the weighted score results and recommend the parking space with the highest score to the user. A path planning unit is configured to generate an optimal navigation path based on the current location of the user and the location information of the recommended parking space.

[0011] Preferably, the underground parking lot management system based on multi-sensor fusion further comprises a heat map generation module configured to generate a parking lot area heat map based on the state determination results of all parking spaces, wherein the heat map generation module comprises: A grid division unit is configured to divide the parking lot area into a plurality of uniform grid units. A real-time heat value calculation unit is configured to calculate a current heat value based on the real-time state data of the parking spaces in each grid unit. A heat value fusion unit is configured to perform weighted fusion of the current heat value and historical heat value to generate a comprehensive heat value of each grid unit. A visualization generation unit is configured to generate a visual heat map based on the comprehensive heat values of all grid units.

[0012] Preferably, the underground parking lot management system based on multi-sensor fusion further comprises an intelligent lighting control module for dynamically adjusting the lighting control strategy of each region according to the activity level of each region in the parking lot region heat map, the intelligent lighting control module comprising: an activity level determination unit for determining the activity level of each grid cell in the parking lot region heat map according to the comprehensive heat value of each grid cell in the parking lot region heat map; a parameter configuration unit for querying a pre-defined lighting control parameter mapping table according to the activity level, and configuring a corresponding lighting control parameter group for each grid cell; the lighting control parameter group comprising a basic brightness value, a vehicle trigger distance, a brightness increase time, and a brightness maintenance time; a brightness increase control unit for controlling the brightness of the lamps and lanterns of the grid cell and adjacent upstream grid cells when detecting that the target enters the vehicle trigger distance of the grid cell, and smoothly adjusting the brightness from the basic brightness value to 100% illumination brightness value according to the brightness increase time; a brightness decrease control unit for maintaining illumination according to the brightness maintenance time after detecting that the target leaves, and then smoothly reducing the brightness to the basic brightness value according to a pre-set brightness decrease curve.

[0013] The second aspect of the present application provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that when the processor executes the computer program, the functions of the underground parking lot management system according to any one of the first aspect are realized.

[0014] The underground parking lot management system based on multi-sensor fusion and the computer device provided by the present application can generate an optimal parking space recommendation and an initial path based on weighted scoring when receiving a request, thereby improving the efficiency of initial guidance and the user parking experience; by continuously acquiring multi-source sensor data after the vehicle enters, full-process monitoring is realized, and the comprehensiveness and real-time performance of parking lot environment and parking space state perception are improved; by predefining a fusion algorithm, multi-source detection data is fused, and the recommended parking space state is determined in real time, thereby improving the accuracy and reliability of parking space state determination; by immediately re-executing the recommendation algorithm and updating the path when determining that the parking space is unavailable, the dynamic adjustment capability and overall service robustness of the system in abnormal conditions are improved.

[0015] Further, the application also realizes vehicle positioning through a Bluetooth beacon network, improves positioning reliability and coverage range; improves the accuracy of target existence and motion judgment through multi-dimensional sensing of parking space state by a radar sensor; improves the accuracy of parking space and vehicle identity recognition through camera and convolutional neural network model identification of visual information; improves the stability and anti-interference ability of positioning data through Kalman filter algorithm processing of Bluetooth signal strength data; improves the accuracy and reliability of vehicle position information through positioning calculation model calculation of vehicle coordinates; improves the uniformity and fusion of information expression through conversion of multi-source data into basic probability distribution; improves the accuracy and anti-interference ability of state judgment decision through Dempster synthesis rule fusion of multi-source evidence; improves the definiteness and operability of state judgment through setting of judgment threshold comparison of comprehensive trust degree; improves the quantitative rationality of distance factor in state discrimination through generation of first probability distribution by vehicle positioning data; improves the discrimination granularity of target existence and motion characteristics through generation of second probability distribution by parking space state sensing data; improves the contribution of image evidence in fusion decision through generation of third probability distribution by visual recognition data; improves the fault tolerance of fusion results to contradictory information through calculation of conflict coefficient between evidence and normalization processing; improves the rationality and accuracy of comprehensive trust degree calculation through synthesis of evidence supporting the same state and normalization; improves the clarity and decision efficiency of state judgment output through extraction of explicit probability value from fused probability distribution; improves the rationality of parking space recommendation and user satisfaction through comprehensive consideration of destination distance, path length, estimated time consumption and user preference for weighted scoring; improves the fine degree of spatial utilization rate perception through division of parking lot into grid units to calculate real-time heat value; improves data stability and trend prediction ability through fusion of historical heat value to generate a comprehensive heat map; improves the monitoring efficiency of parking lot operation state through visual presentation of regional heat distribution; improves the fine and energy efficiency of lighting control through differential configuration of lighting parameters according to heat map activity level differences; improves driving safety and guidance through pre-lighting of lamps in the current area and upstream area triggered by the vehicle; improves visual comfort and reduces energy consumption through smooth adjustment of brightness and maintenance time design.

[0016] In summary, the underground parking lot management system and computer equipment based on multi-sensor fusion proposed by the application solve the technical problem that the prior art cannot realize dynamic guidance adjustment when the parking space state is abnormal due to inaccurate perception and decision lag. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments 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 described below only show some of the embodiments of the present application, and are not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0018] Figure 1 The module schematic diagram of the underground parking lot management system based on multi-sensor fusion provided by an embodiment of the present application is shown in the figure. Figure 2 The unit schematic diagram of the data acquisition module provided by an embodiment of the present application is shown in the figure. Figure 3 The unit schematic diagram of the positioning data acquisition unit provided by an embodiment of the present application is shown in the figure. Figure 4 The unit schematic diagram of the fusion determination module provided by an embodiment of the present application is shown in the figure. Figure 5 The unit schematic diagram of the probability distribution generation unit provided by an embodiment of the present application is shown in the figure. Figure 6 The unit schematic diagram of the evidence fusion unit provided by an embodiment of the present application is shown in the figure. Figure 7 The unit schematic diagram of the parking space recommendation and navigation module provided by an embodiment of the present application is shown in the figure. Figure 8 The module schematic diagram of another underground parking lot management system based on multi-sensor fusion provided by an embodiment of the present application is shown in the figure. Figure 9 The unit schematic diagram of the heat map generation module provided by an embodiment of the present application is shown in the figure. Figure 10 The unit schematic diagram of the intelligent lighting control module provided by an embodiment of the present application is shown in the figure. Figure 11 The schematic diagram of the computer device provided by an embodiment of the present application is shown in the figure.

[0019] The implementation of the present application, functional features and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] It should be noted that if the embodiments of the present application involve directional indications, such as up, down, left, right, front, back, etc., the directional indications are only used to explain the relative position relationship, motion condition, etc. between the components in a certain posture, and if the certain posture changes, the directional indications will also change accordingly.

[0022] In addition, if the embodiments of the present application involve descriptions such as "first", "second", etc., the descriptions of "first", "second", etc. are only for description purposes and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first" and "second" can be explicitly or implicitly included at least one of the features. In addition, the meaning of "and / or" appearing throughout the text includes three parallel schemes. For example, "A and / or B" includes A scheme, or B scheme, or A and B scheme. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of ordinary skilled in the art, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor within the scope of protection required by the present application.

[0023] The main purpose of the present application is to provide a multi-sensor fusion based underground parking lot management method and system, which aims to solve the technical problems that the existing technology cannot realize dynamic guidance adjustment when the parking space state is abnormal due to inaccurate perception and decision lag.

[0024] The system hardware architecture mainly includes a Bluetooth Mesh network module, a radar detection module, a camera recognition module, an intelligent gateway and a cloud server deployed in the parking lot, and the software aspect includes a user terminal APP. The Bluetooth Mesh network is composed of a large number of Bluetooth beacon nodes distributed inside the parking lot, forming a communication and positioning network covering the entire parking area. The radar sensor is installed above each parking space to detect the existence and motion state of objects within the parking space. The high-definition camera module is deployed above the key nodes of the lane and the parking space group, responsible for collecting video stream data. The intelligent gateway is an edge computing node, responsible for collecting and preprocessing data from Bluetooth, radar and camera. The cloud server is the core brain of the system, running core algorithms such as parking space management, data fusion, navigation scheduling, etc. Users interact with the system through mobile APP or applet.

[0025] As shown in Figures 1 to 11 The first aspect of the present application proposes a multi-sensor fusion based underground parking lot management system, which comprises: A parking space recommendation and navigation module is used to receive user parking requests, generate parking space recommendation results based on preset evaluation indicators through a weighted scoring algorithm, and generate an initial navigation path from the user's current location to the recommended parking space; The data acquisition module is configured to, in response to detecting that a user vehicle enters a preset monitoring area of a parking lot, continuously acquire multi-source detection data from multiple types of sensors, the multi-source detection data including vehicle positioning data, parking space state perception data, and visual recognition data. The fusion determination module is configured to fuse the multi-source detection data according to a preset parking space data fusion algorithm, and determine the state of the recommended parking space in real time. The dynamic scheduling module is configured to, in response to the state of the recommended parking space becoming unusable, re-execute the weighted scoring algorithm based on real-time multi-source detection data and preset evaluation indexes, generate a new parking space recommendation result, and update a navigation path.

[0026] Specifically, referring to Figure 1In a specific embodiment of the present application, the parking lot management system includes a parking space recommendation and navigation module, a data acquisition module, a fusion determination module, and a dynamic scheduling module. When the vehicle owner approaches the destination and needs to park, the parking request can be submitted through the mobile phone APP or applet, and the request includes the destination information and personal parking preference such as proximity to elevator or charging pile. After the parking space recommendation and navigation module of the system receives the request, it first obtains all real-time idle parking space information in the parking lot to form a set of available parking spaces. Through a weighted scoring algorithm, each parking space in the set is calculated based on multiple evaluation indicators. The algorithm considers multiple evaluation indicators, including the straight-line distance from the parking space to the user's destination, the optimal path distance from the user's current location to the parking space, the estimated driving time, and the matching degree of the parking space characteristics and user preference. Each indicator is standardized to a percentage score and assigned a different weight, for example, the closest to the destination has the highest weight, and the preference matching degree has a relatively low weight. By calculating the weighted total score of each parking space and sorting it from high to low, the highest scoring parking space is finally locked for a period of time, and an optimal driving navigation path from the user's current location to the recommended parking space is generated for the user. When the user's vehicle has entered the internal monitoring area of the parking lot detected by the entrance gate, ground coil, or entrance camera, the multi-sensor data acquisition process is immediately activated. The data acquisition module continuously acquires multi-source data from different types of sensors, including real-time positioning data of the vehicle estimated by the user's mobile phone APP through the received Bluetooth Mesh network signal strength deployed inside the parking lot, parking space state perception data collected by the radar sensor installed above each parking space (judging whether there is a vehicle or other object on the parking space and its motion state), and visual recognition data captured by the camera module deployed at key nodes and processed by the image recognition model (used to confirm the parking space number and identify the license plate information of the parked vehicle). Then the fusion determination module calls the preset parking space data fusion algorithm to fuse and process the real-time acquired vehicle positioning data, parking space state perception data, and visual recognition data, and calculates the comprehensive determination result of the current state of the recommended parking space. If the fusion determination result confirms that the recommended parking space is still available, the user will be guided to the new target parking space. If the recommended parking space status becomes unavailable due to other vehicles parking first or system recognition errors during the guidance process, the system will immediately trigger the dynamic scheduling mechanism. The dynamic scheduling module will re-evaluate the new optimal recommended parking space from the current actual available parking space set based on the latest global parking lot space state data, and immediately update the navigation path for the user to guide them to the new target parking space, ensuring that the user can finally park successfully.

[0027] It can be understood that the embodiment improves the accuracy and reliability of the parking space state determination by multi-sensor data fusion and real-time analysis, reduces misjudgment and missed judgment; improves the parking space recommendation efficiency and user parking experience by intelligent weighted scoring and dynamic rescheduling, reduces the time of users searching for parking spaces and the invalid driving of vehicles in the field; improves the robustness and fault tolerance of the navigation system by real-time data driven path updating, avoids guiding the user to the occupied parking space.

[0028] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, such as replacing the Bluetooth Mesh network with a UWB ultra-wideband positioning system to improve positioning accuracy; or replacing the radar detection module with an ultrasonic sensor to adapt to different environmental detection requirements; or replacing the weighted average scoring algorithm with a fuzzy logic-based evaluation method to handle more complex uncertainty preferences; or replacing the preset fixed weight evaluation index with an adaptive algorithm that dynamically adjusts the weight according to real-time traffic flow or historical user behavior data to enhance system flexibility.

[0029] Preferably, the data acquisition module comprises: A data triggering unit is configured to trigger a data acquisition process when detecting that a user's vehicle enters the preset monitoring area; A positioning data acquisition unit is configured to acquire vehicle positioning data reported by a user terminal; the user terminal receives signals of multiple Bluetooth beacons in a parking lot and calculates the vehicle positioning data based on a positioning algorithm; A radar data acquisition unit is configured to acquire parking space state perception data through a radar sensor deployed above a parking space; the parking space state perception data includes the existence state, motion state, and contour feature information of a target object on the parking space; A visual data acquisition unit is configured to acquire visual data through a camera module deployed in the parking lot, and generate visual recognition data including parking space number and license plate number recognition results by using a target detection model based on a convolutional neural network to perform image recognition processing on the visual data.

[0030] Specifically, referring to Figure 2In a specific embodiment of the present application, the data acquisition module includes a data trigger unit, a positioning data acquisition unit, a radar data acquisition unit, and a visual data acquisition unit. When the data trigger unit detects that the user's vehicle enters the underground parking lot entrance and triggers the ground inductor, or the automatic license plate recognition system recognizes the user's vehicle, it is determined that the user's vehicle enters the preset monitoring area, and the system immediately activates the data acquisition process. First, the positioning data acquisition unit uses the special application on the user's mobile phone as the user terminal to start scanning and receiving the signals broadcast by multiple Bluetooth beacons pre-deployed on the parking lot pillars or ceiling. Based on the positioning algorithm running inside the application, such as the received signal strength indication-based triangulation method, the distance between the terminal and each beacon is calculated in real time, and the rough coordinate position of the vehicle in the current parking lot map is calculated, and the vehicle positioning data is packaged and reported to the system server. At the same time, the radar data acquisition unit continuously scans and detects through the radar sensor installed above each parking space. In one possible embodiment, a 24G millimeter wave radar sensor is used for scanning and detection, which can emit electromagnetic waves and receive echoes. By analyzing the characteristic changes of the echo signals, it can accurately determine whether a vehicle is parked on the parking space, and can also perceive more subtle target information, such as distinguishing between a vehicle and a person's temporary stay, even capturing the target's motion state such as stationary or slow movement, and obtaining the approximate outline features of the target, thereby forming rich parking space state perception data. At the same time, the visual data acquisition unit continuously collects video streams through the camera module deployed above the key passageway of the parking lot or directly facing the parking area. The pre-trained target detection model based on convolutional neural network, such as YOLO or SSD model, performs real-time image analysis to identify each parking area in the video frame and extract the printed parking number digits in that area. At the same time, if a vehicle is parked on the parking space, the model will also locate and identify the front or rear license plate number of the vehicle. Finally, the structured data generated after analyzing each frame of image, which contains the accurate parking number and the possible license plate number recognition result, is uploaded as visual recognition data. The parking lot management system continuously receives and caches the Bluetooth positioning data from the user terminal, the state perception data from each parking radar, and the visual recognition data from the camera, providing sufficient data for subsequent multi-source data fusion decision-making.

[0031] It can be understood that the present embodiment improves the diversity and complementarity of information sources through the cooperative data acquisition of three types of heterogeneous sensors, namely Bluetooth beacon network, millimeter wave radar, and camera, laying a solid foundation for subsequent fusion decision-making. Through the multi-dimensional perception of the radar on the existence, motion, and outline of the target, the precision and environmental interference resistance of the parking space state judgment are improved. Through real-time analysis of the video stream by the convolutional neural network model, the automation and accuracy of parking number and license plate number recognition are improved.

[0032] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, such as replacing the Bluetooth beacon with a ZigBee node or a LoRa beacon to build different types of indoor positioning networks; or replacing the target detection model based on the convolutional neural network with a visual recognition model DETR based on the Transformer architecture to improve the ability to capture long-distance dependencies; or allowing the user terminal to perform auxiliary positioning through a Wi-Fi probe or UWB ultra-wideband technology as a supplement or alternative to Bluetooth positioning.

[0033] Preferably, the positioning data acquisition unit comprises: a signal receiving subunit for controlling the user terminal to receive signal strength data from multiple Bluetooth beacons; a filtering processing subunit for processing the signal strength data using a Kalman filtering algorithm to filter out noise and improve positioning stability; a positioning solution subunit for calculating the vehicle position coordinates based on the processed signal strength data through a positioning solution model to generate the vehicle positioning data.

[0034] Specifically, referring to Figure 3In a specific embodiment of the present application, the positioning data acquisition unit comprises a signal receiving subunit, a filtering processing subunit and a positioning calculation subunit. The signal receiving subunit controls the application software running on the user's mobile phone to continuously scan the surrounding environment after the vehicle enters the parking lot, search and receive the wireless signals broadcasted by the multiple Bluetooth beacons fixedly installed at different positions inside the parking lot. First, the unique identifier corresponding to each detected Bluetooth beacon is read, and the strength RSSI value of the received beacon signal is measured, thereby obtaining a set of original signal strength data from different beacons. Due to the complex environment of the underground parking lot, there are multipath effects, human shielding and other interferences, which cause the original RSSI value to fluctuate greatly. In order to improve the data quality, the Kalman filtering algorithm is applied to process the batch of original signal strength data by the filtering processing subunit, which specifically includes: first, a Kalman filter instance is individually initialized for each Bluetooth beacon to be tracked. The internal state of the filter includes two key quantities: one is the instantaneous dynamic distance estimation value between the mobile terminal and the Bluetooth beacon, and the other is the rate of change of the distance over time (i.e. radial velocity). According to the prior knowledge of the typical motion mode of the vehicle in the parking lot, the process noise parameters and observation noise parameters are configured for the filter, which reflect the assessment of the uncertainty of the vehicle motion model and the error characteristics of the Bluetooth chip RSSI measurement. The processing is performed in discrete time steps. In the prediction stage of each time step, the filter uses the distance and speed optimally estimated at the last time to calculate the predicted value of the distance and speed at the current time based on a constant speed dynamic model, and updates the uncertainty of the state estimation. In the update stage, the newly collected original RSSI measurement value is sent to the filter. According to the classic wireless signal path loss model, the RSSI value is converted into an equivalent distance observation value. The difference between the newly obtained distance observation value and the predicted distance value obtained in the prediction stage is calculated, and the Kalman gain parameter is calculated in real time. The gain parameter dynamically balances the trust degree of the system dynamic model prediction value and the trust degree of the current fresh observation value. Finally, the predicted distance and speed states are weighted and corrected using the Kalman gain to produce the posterior state estimation at the current time, which is closest to the optimized real situation, and the optimized distance estimation value is output for subsequent positioning calculation. The filtering process is independently and continuously performed on the RSSI data stream of each beacon, which can effectively smooth the random spikes and short-term fluctuations in the original signal sequence, and output a stable and trend-smooth distance change curve. After obtaining the estimated distance from the terminal to at least three Bluetooth beacons with known coordinate positions, the positioning calculation subunit calculates the specific position coordinates of the user terminal / vehicle in the two-dimensional plane of the parking lot by using mathematical methods such as trilateration positioning method through a positioning calculation model. Finally, the system generates a vehicle positioning data packet containing the coordinate information and its timestamp, which is used for subsequent data fusion and parking space state determination.

[0035] It can be understood that the embodiment smoothes and denoises the original Bluetooth signal strength data and estimates the data by the Kalman filtering algorithm, improves the stability and reliability of the positioning data input, drives the positioning solution model by the processed clean data, improves the calculation accuracy of the final vehicle position coordinates, improves the data quality of the positioning link, provides more accurate position basis for subsequent multi-sensor fusion decision, and indirectly improves the robustness of the entire parking space state determination system.

[0036] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, for example, replacing the Kalman filtering algorithm with a particle filtering algorithm to better handle non-Gaussian noise of the signal strength, or replacing the trilateration method relied on in the positioning solution model with a fingerprint positioning method to utilize the pre-collected signal strength map for matching.

[0037] Preferably, the fusion determination module comprises: a probability distribution generation unit configured to generate a plurality of basic probability distributions corresponding to the parking space occupancy states based on the vehicle positioning data, the parking space state perception data and the visual recognition data; an evidence fusion unit configured to perform fusion calculation on the plurality of basic probability distributions based on a Dempster combination rule to obtain a comprehensive trust degree of the recommended parking space being occupied; a state determination unit configured to compare the comprehensive trust degree with a preset determination threshold, and determine the state of the recommended parking space according to a comparison result.

[0038] Specifically, referring to Figure 4In a specific embodiment of the present application, the fusion determination module comprises a probability assignment generation unit, an evidence fusion unit and a state determination unit; wherein, the system executes a preset parking space data fusion algorithm, and the process of determining the state of the recommended parking space is as follows: the probability assignment generation unit first generates independent basic probability assignments based on the real-time acquired multi-source detection data, including vehicle positioning data representing the macro position of the vehicle, parking space state perception data reflecting the physical space state of the parking space, and visual recognition data providing accurate visual evidence; for each piece of data, a basic probability assignment is calculated according to a preset perception model, a confidence algorithm and a historical accuracy rate through a preset recognition framework (including possible states of the parking space such as occupied and idle), which quantifies the support degree of the evidence of the sensor to each single point hypothesis in the recognition framework and the recognition framework itself; then, the evidence fusion unit uses the Dempster combination rule as the core fusion operator, takes the above-mentioned multiple basic probability assignments as input, calculates the joint support degree between each evidence body and processes possible evidence conflicts, performs orthogonal sum operation, and combines the evidence two by two in turn, each combination produces a new synthesized evidence body that has fused more information, until all evidences are merged, and the process finally outputs a comprehensive basic probability assignment containing the comprehensive confidence degree of the recommended parking space being occupied; finally, the state determination unit compares the calculated comprehensive confidence degree with a preset judgment threshold value through experiment or experience; if the comprehensive confidence degree is higher than the threshold value, the recommended parking space is determined to be in the occupied state; if it is lower than the threshold value, it may be determined to be in the idle state or marked as an uncertain state due to evidence conflict or insufficiency, thereby triggering the corresponding subsequent logic.

[0039] It can be understood that, by converting multi-source heterogeneous data into a unified basic probability assignment form and performing mathematical fusion, the embodiment improves the processing capability of the state determination process for uncertain information and the scientificity of the decision; by using the Dempster combination rule to effectively synthesize the advantages of each sensor evidence and resolve conflicts, the accuracy and reliability of the final state determination result are improved; by setting a judgment threshold value to convert the continuous confidence degree into a discrete state decision, the definiteness and operability of the system decision are improved.

[0040] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, for example, replacing the Dempster combination rule with other evidence theory combination rules such as the Yager rule or Murphy's simple average method to change the processing strategy of conflicting evidence; or discounting the basic probability assignment before fusion to adjust the evidence weight according to the sensor reliability; or using fuzzy set theory to convert the sensor data into a membership function and then perform fusion to handle the semantic level uncertainty.

[0041] Preferably, the probability distribution generation unit comprises: a first distribution sub-unit configured to calculate a distance relationship between a user vehicle and the recommended parking space based on the vehicle positioning data, and generate a first basic probability distribution; a second distribution sub-unit configured to generate a second basic probability distribution based on a target presence state and a motion feature in the parking space state perception data; a third distribution sub-unit configured to generate a third basic probability distribution based on a parking space number and a license plate number recognition result in the visual recognition data.

[0042] Specifically, referring to Figure 5In a specific embodiment of the present application, the probability distribution generation unit includes a first distribution sub-unit, a second distribution sub-unit and a third distribution sub-unit; first, generate the corresponding basic probability distribution for three different types of detection data respectively: for vehicle positioning data, the first distribution sub-unit calculates the two-dimensional plane distance between the filtered vehicle coordinates reported by the user terminal and the center point of the recommended parking space to be determined, and queries the pre-set distance-probability mapping relationship table according to the calculated distance value; the mapping relationship table is obtained based on historical data statistics, and the core logic is that the closer the distance, the higher the possibility of the vehicle parking in the parking space, that is, the higher the basic probability mass assigned to the "occupied" state, and the farther the distance, the higher the probability mass assigned to the "idle" state; at the same time, no matter how far the distance is, a small part of the probability mass will be assigned to the "uncertain" universal proposition to represent the inherent error of the positioning system itself and the uncertainty introduced by the transition state that the vehicle may be driving to or away from the parking space, thereby generating the first basic probability distribution. For parking space state sensing data, the second distribution sub-unit analyzes the original detection results from the millimeter wave radar, including target existence confidence and target motion characteristic parameters, and the second distribution sub-unit comprehensively evaluates these parameters: if the radar reports a high-confidence target existence and the motion characteristic indicates a stationary state, a basic probability distribution highly supporting the "occupied" state is generated; if the radar reports no target or a very low target confidence, a basic probability distribution highly supporting the "idle" state is generated; if the target exists but is in a significant motion state, it may indicate that a pedestrian or a shopping cart is passing by, at which time the system will reduce the support for any one of the "occupied" or "idle" single-point propositions, and assign more probability mass to the "uncertain" proposition, indicating that the current radar evidence cannot clearly determine the parking space state, thereby generating the second basic probability distribution. For visual recognition data, the third distribution sub-unit analyzes the structured results output by the camera module through the image recognition model, and the core is the parking space number recognition confidence and the license plate number recognition result; if the system recognizes the correct parking space number and at the same time recognizes the valid license plate number, a basic probability distribution strongly supporting the "occupied" state is generated; if the system recognizes the correct parking space number but does not recognize the license plate (may be due to the parking space being idle, or due to obstruction, light, etc. Recognition failure), a basic probability distribution tending to support the "idle" state is generated; if the parking space number recognition itself has low confidence, the system will assign most of the probability mass to the "uncertain" proposition, indicating that the visual evidence itself is unreliable, thereby generating the third basic probability distribution. The three basic probability distributions are collectively prepared for the subsequent evidence fusion step.

[0043] It can be understood that the embodiment generates the first basic probability assignment by establishing a distance-probability mapping model for vehicle positioning data, converts a continuous physical distance quantity into a quantitative trust degree for the parking space state, improves the rationality of state discrimination using positioning information, generates the second basic probability assignment by comprehensively evaluating the radar target existence and motion characteristics, enables the system to distinguish between different scenarios such as vehicle static occupation and temporary object passing, and improves the state discrimination accuracy of radar data, and generates the third basic probability assignment by analyzing the parking space number and license plate number information in the visual recognition result, converts the classification result of image recognition into a probability evidence, and improves the contribution and reliability of visual evidence in fusion.

[0044] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, for example, when generating the first basic probability assignment, replacing the simple distance-probability mapping table with a continuous probability distribution model based on a Gaussian function to more smoothly reflect distance uncertainty; or when generating the second basic probability assignment, in addition to the target existence and motion state, further introducing a radar point cloud contour matching degree as another dimension of generating probability; or when generating the third basic probability assignment, subdividing the effectiveness of license plate recognition results into multiple grades such as high confidence recognition, low confidence recognition, and recognition failure, and assigning different probability assignment strategies respectively.

[0045] Preferably, the evidence fusion unit comprises: A conflict calculation subunit is configured to calculate a conflict coefficient between the first basic probability assignment, the second basic probability assignment, and the third basic probability assignment. A synthesized evidence generation subunit is configured to synthesize all evidences in the plurality of basic probability assignments supporting the "recommended parking space being occupied" state to obtain a preliminary synthesized evidence, and calculate a normalization factor using the conflict coefficient. An integrated probability assignment subunit is configured to divide the preliminary synthesized evidence by the normalization factor to obtain a fused integrated basic probability assignment. A trust degree extraction subunit is configured to extract a probability value corresponding to the "recommended parking space being occupied" state from the integrated basic probability assignment as the integrated trust degree.

[0046] Specifically, referring to Figure 6In a specific embodiment of the present application, the evidence fusion unit comprises a conflict calculation subunit, a synthetic evidence generation subunit, a comprehensive probability distribution subunit and a trust degree extraction subunit. After obtaining three basic probability distributions respectively corresponding to the vehicle positioning data, the parking space state perception data and the visual recognition data, the Dempster combination rule is executed to start the fusion calculation. First, the conflict calculation subunit combines the three pieces of evidence two by two, for example, the first basic probability distribution is combined with the second basic probability distribution, the sum of the products of all evidence combinations that can support the same hypothesis is calculated, and specifically, for the hypothesis that the recommended parking space is occupied, the system finds the probability value supporting occupation in the first distribution and the probability value supporting occupation in the second distribution, and multiplies them. At the same time, the evidence supporting uncertainty is also considered, for example, the probability value supporting occupation in the first distribution is multiplied by the probability value assigned to the uncertainty set in the second distribution, and the probability value assigned to the uncertainty set in the first distribution is multiplied by the probability value supporting occupation in the second distribution. The sum of these products is added to obtain the joint probability mass of the preliminary support for the occupation hypothesis after the fusion of the first and second evidence. Then, the synthetic evidence generation subunit processes the conflict between the evidence, that is, the sum of the probability products of the mutually exclusive cases that the first distribution supports occupation and the second distribution supports vacancy, and the first distribution supports vacancy and the second distribution supports occupation, is calculated as the conflict coefficient between the two pieces of evidence, and then the conflict coefficient is used to calculate the normalization factor. Subsequently, the comprehensive probability distribution subunit divides the joint probability mass of the preliminary support for occupation by the normalization factor to obtain the probability mass of the new evidence body after the fusion of the first and second evidence for the occupation hypothesis, so that the new evidence body contains the information of the two sensors. Then the same combination operation is performed on the new evidence body and the third basic probability distribution: the joint probability mass of the new evidence body and the third evidence supporting occupation is calculated, the new conflict coefficient between them is calculated, and normalization processing is performed to finally obtain a comprehensive basic probability distribution that fuses all three sources of evidence. Finally, the trust degree extraction subunit extracts the probability value definitely assigned to the proposition that the recommended parking space is occupied from the final comprehensive basic probability distribution as the comprehensive trust degree for determining the parking space state.

[0047] It can be understood that the embodiment ensures the mathematical standard processing of multi-source evidence, improves the scientificity and repeatability of the fusion process by performing two-by-two progressive evidence fusion based on the Dempster combination rule. By calculating and using the conflict coefficient for normalization, the negative impact of mutually contradictory evidence on the fusion result is effectively reduced, and the rationality of the final comprehensive trust degree is improved. By accurately extracting the probability value corresponding to the "occupied" state from the final comprehensive basic probability distribution as the output, a clear and quantitative basis is provided for the subsequent threshold comparison decision.

[0048] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, for example, when the conflict coefficient is greater than the preset threshold, instead of immediately synthesizing, an instruction for re-acquiring data is triggered; or changing the order of evidence synthesis, for example, first fusing visual and radar evidence and then fusing positioning evidence; or using a parallel fusion architecture to process multiple evidences at the same time instead of serialized two-by-two fusion.

[0049] Preferably, the parking space recommendation and navigation module comprises: A request receiving unit configured to receive a parking request submitted by a user through a mobile terminal, the request comprising destination information and a parking space preference type; A parking space screening unit configured to obtain real-time parking space state data of a parking lot and screen out a set of available parking spaces; A score calculation unit configured to calculate a weighted score for each parking space in the set of available parking spaces based on a distance to the destination, a path planning distance, an estimated time consumption, and a parking space preference matching degree; A recommendation generation unit configured to sort the parking spaces according to the weighted score results and recommend the parking space with the highest score to the user; A path planning unit configured to generate an optimal navigation path based on the current location of the user and the location information of the recommended parking space.

[0050] Specifically, referring to Figure 7In a specific embodiment of the present application, the parking space recommendation and navigation module includes a request receiving unit, a parking space screening unit, a score calculation unit, a recommendation generating unit and a path planning unit; a user submits a parking request through an application on a smartphone, specifies a final destination (e.g. the third elevator hall of a certain shopping mall), and selects a parking space preference type (e.g. hopes to park in a parking space close to the elevator); the request receiving unit immediately queries the real-time data platform of the parking lot after receiving the request, obtains the current state information of all parking spaces through the parking space screening unit, filters out those parking spaces that have been occupied or have been reserved by other users, and forms an initial available parking space set; then the score calculation unit calculates a weighted score for each parking space in the available parking space set based on the following four evaluation indicators: the first indicator is the straight-line distance of the parking space from the destination, calculating the straight-line length from the center point of each available parking space to the third elevator hall specified by the user; the second indicator is the path planning distance, calculating the actual length of the optimal path from the entrance of the parking lot to the parking space according to the electronic map and traffic rules inside the parking lot; the third indicator is the estimated time consumption, the system estimates the approximate time required to reach the parking space from the entrance by combining the path distance and the typical driving speed inside the parking lot; the fourth indicator is the parking space preference matching degree, the system checks whether the attributes of the parking space match the "close to the elevator" preference selected by the user, and gives a high score for complete compliance, and the score decreases for partial compliance or non-compliance; the value of each indicator is converted to a standardized score between 0 and 100 through the min-max normalization method, and then the comprehensive score of each parking space is calculated, the calculation method is: multiplying the standardized score of the distance from the destination by a weight of 5, multiplying the score of the path planning distance by a weight of 2, multiplying the score of the estimated time consumption by a weight of 2, multiplying the score of the parking space preference matching degree by a weight of 1, and finally adding the four weighted scores to obtain the final weighted total score of the parking space. Then the recommendation generating unit sorts the weighted total scores of all available parking spaces from high to low, selects the parking space with the highest total score as the final recommendation result, and notifies the user through the application interface, while marking the parking space status as reserved; finally, the path planning unit calls the path planning engine to generate a specific and directionally navigable optimal driving path based on the real-time current location provided by the user's smartphone GPS and the coordinates of the recommended parking space, and sends the initial navigation path to the user's mobile terminal application to guide the user to the recommended parking space. It should be noted that the weights for calculating the comprehensive score of each parking space can be adjusted as needed, and the weights in this embodiment are only for calculation examples.

[0051] It can be understood that the embodiment improves the comprehensive optimality and personalization degree of the parking space recommendation result by comprehensively considering and weighting multi-dimensional indexes such as distance, path, time consumption and user preference; the comparability between different parking spaces and the fairness of the scoring algorithm are improved by standardizing and weighting summing the indexes of different dimensions; the efficiency and user experience of the entire parking guidance process are improved by generating recommendations and navigation based on real-time parking space status and dynamic path planning.

[0052] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, for example, replacing the weighted scoring algorithm with a decision model based on the analytic hierarchy process to more structurally determine the index weight; or replacing one of the four evaluation indexes with the real-time congestion degree or historical safety record of the area where the parking space is located; or replacing the simple linear weighted sum with the TOPSIS multi-attribute decision method based on ideal point ranking.

[0053] Preferably, the underground parking lot management system based on multi-sensor fusion further comprises a heat map generation module for generating a parking lot area heat map based on the state determination results of all parking spaces, the heat map generation module comprising: a grid division unit for dividing the parking lot area into a plurality of uniform grid units; a real-time heat value calculation unit for calculating a current heat value based on the real-time state data of the parking spaces in each grid unit; a heat value fusion unit for weighting and fusing the current heat value and the historical heat value to generate a comprehensive heat value of each grid unit; a visualization generation unit for generating a visual heat map based on the comprehensive heat values of all grid units.

[0054] Specifically, refer to Figure 8 and Figure 9In a specific embodiment of the present application, the heat map generation module comprises a grid division unit, a real-time heat value calculation unit, a heat value fusion unit and a visualization generation unit. First, the grid division unit divides the electronic map of the entire parking lot into a plurality of square grid units of uniform size, each grid unit covering a certain physical area (for example, one square meter), and each grid unit is assigned a unique coordinate identifier. Then, based on the real-time parking space state determination results obtained by the fusion determination module (i.e., knowing whether each parking space is currently in an occupied state or an idle state), the real-time heat value calculation unit calculates the current heat value of each grid unit. The calculation logic is as follows: count all parking spaces falling within the boundaries of the grid unit, and calculate the ratio of the number of parking spaces in the occupied state to the total number of parking spaces in the grid unit. This ratio is taken as the current heat value reflecting the current instantaneous usage density of the grid unit. Subsequently, the heat value fusion unit introduces the time dimension and performs data fusion. The historical heat values of the grid unit in the same period in the recent past (for example, the average heat value in the same hour in the past seven days) are retrieved from the storage. A weighted fusion algorithm is used to combine the calculated current heat value with the retrieved historical heat value. Specifically, a higher weight is assigned to the current heat value, for example, 0.7, and a lower weight is assigned to the historical heat value, for example, 0.3. Then, the weighted values are added to generate a comprehensive heat value that reflects both real-time conditions and historical patterns. The weighted fusion algorithm is performed for all grid units one by one. Finally, the visualization generation unit generates a visual heat map based on the set of comprehensive heat values of all grid units using image rendering technology. The grid units with different comprehensive heat value ranges on the heat map are filled with different colors. A gradient color system from cool tones to warm tones is usually used, for example, blue represents low usage and red represents high usage, thereby intuitively showing the spatial usage heat distribution of different areas in the parking lot, which is used for macroscopic monitoring by system administrators or as input data for other optimization modules.

[0055] It can be understood that the present embodiment improves the degree of refinement and real-time of the perception of parking space utilization by dividing the parking lot into grids and calculating the current heat value based on real-time parking space status. By weighting and fusing the current heat value with the historical heat value, the generated heat map can reflect the immediate situation and smooth accidental fluctuations, improving the stability and trend prediction ability of the data. Based on the comprehensive heat value, a visual heat map is generated to improve the intuitiveness and understandability of the overall operation state of the parking lot, assisting macro management decisions.

[0056] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, such as replacing the simple fixed-size grid division with irregular area division according to the parking space layout or passage structure; or when calculating the current heat value, not only considering the number of parking space occupancy, but also introducing the average occupancy time of parking space as a calculation factor; or replacing the fixed historical weight with an adaptive weight dynamically adjusted according to whether the date type is a weekday or a holiday.

[0057] Preferably, the underground parking lot management system based on multi-sensor fusion further comprises an intelligent lighting control module for dynamically adjusting the lighting control strategy of the corresponding area according to the activity level of each area in the parking lot area heat map, the intelligent lighting control module comprising: an activity level determination unit for determining the corresponding activity level according to the comprehensive heat value of each grid unit in the parking lot area heat map; a parameter configuration unit for querying a predefined lighting control parameter mapping table according to the activity level and configuring a corresponding lighting control parameter group for each grid unit; the lighting control parameter group comprising a basic brightness value, a vehicle trigger distance, a brightness boost time, and a brightness maintenance time; a brightness boost control unit for controlling the brightness of the lamps and lanterns of the grid unit and adjacent upstream grid units when detecting that the target enters the vehicle trigger distance of the grid unit, and smoothly adjusting the brightness from the basic brightness value to 100% illumination brightness value according to the brightness boost time; a brightness drop control unit for maintaining illumination according to the brightness maintenance time after detecting that the target leaves, and then smoothly reducing the brightness to the basic brightness value according to a preset brightness drop curve.

[0058] Specifically, see Figure 8 and Figure 10In a specific embodiment of the present application, the intelligent lighting control module includes an activity determination unit, a parameter configuration unit, a brightness increase control unit and a brightness decrease control unit. First, through the activity determination unit, the latest generated parking lot area heat map is read, each grid cell in the map contains a comprehensive heat value, according to the size of the comprehensive heat value of each grid cell, an activity level is determined for each grid cell, for example: the top 15% area with the highest heat value is divided into a high activity level, the next 10% area is divided into a medium activity level, and the remaining area is divided into a low activity level. After determining the activity level, the parameter configuration unit queries the pre-defined lighting control parameter mapping table to obtain the lighting control parameters corresponding to each activity level, for example, for the grid cell of the high activity level, the corresponding parameter group is: the basic brightness value is 20%, the vehicle triggering distance is 15 meters, the brightness increase time is 4 seconds, and the brightness maintenance time is 18 seconds; for the grid cells of the medium activity level and the low activity level, different parameter values are configured respectively, for example, the basic brightness of the medium activity area is lower, the triggering distance is shorter, and the brightness change time is longer. Then, these parameter groups are sent to the intelligent lighting controller associated with the corresponding grid cell, each controller cooperates with the luminaires it controls and a 24G radar sensor to work, when the radar sensor detects that a vehicle enters the preset vehicle triggering distance of the grid cell it belongs to, for example, in the high activity area, it is detected that a vehicle approaches within 15 meters, the brightness increase control unit controls the lighting controller of the grid cell and the adjacent upstream grid cell, which is determined according to the vehicle driving direction, the brightness increase control unit controls the brightness of the LED luminaires, according to the brightness increase time set in the parameter group, the brightness is smoothly and non-step increased from the basic brightness value to 100% full lighting brightness; when the radar sensor detects that the vehicle has left the area, the brightness decrease control unit continues to maintain 100% brightness for a period of time according to the brightness maintenance time set in the parameter group, and then controls the brightness to gradually decrease to the initially set basic brightness value according to a preset smooth decrease curve, completing a complete lighting control cycle. The lighting of different areas of the entire parking lot is different according to the actual use heat, and a differentiated, demand-matched energy-saving lighting strategy is implemented.

[0059] It can be understood that, in the embodiment, different lighting parameters are configured for different areas through the activity level defined based on the heat map, the fine on-demand allocation of lighting resources is realized, and the overall energy efficiency is improved; through the radar detection outside the vehicle triggering distance, the luminaires in the current area and the upstream area are lit in advance, and a smooth brightness adjustment curve is adopted, the user experience and safety are improved; by setting the brightness maintenance time, the frequent flickering of light caused by the temporary stay of the vehicle is avoided, and the stability of the lighting control and the service life of the equipment are improved.

[0060] Based on the above technical solutions, those skilled in the art can make corresponding equivalent improvements according to the specific characteristics of the application scene or system requirements, such as subdividing the activity level from three levels to five levels or mapping the parameters in a continuous value manner to achieve a smoother control gradient; or replacing the fixed brightness maintenance time with a dynamic time length associated with the actual stay time of the vehicle detected by the radar; or introducing ambient light intensity sensor data as a compensation factor in lighting control to further reduce the basic brightness in areas with sufficient natural light.

[0061] Further, the multi-sensor fusion-based underground parking lot management system provided by the application further includes a reverse car search function. When a user needs to leave the parking lot after completing shopping or business, a reverse car search request can be initiated through a special application or applet installed on a smartphone. After receiving the request, the system obtains the user's current location information through the application, uses the built-in GPS module of the mobile phone, the Bluetooth positioning base station deployed inside the parking lot, or the Wi-Fi signal for hybrid positioning, and calculates the precise coordinates of the user in the parking lot. The system then calls a path planning algorithm, plans an optimal walking path from the user's current location to the vehicle parking location based on the electronic map data of the parking lot and in combination with real-time congestion information, such as the activity level of each area obtained from the heat map analysis. The path planning algorithm considers factors such as shortest distance, least turns, and avoidance of high congestion areas, and uses a dynamic updating mechanism to immediately recalculate the path when temporary obstacles or congestion changes are detected on the path. The navigation information is displayed in a graphical manner through the application interface, providing turn-by-turn directions and supporting voice prompts to enhance user experience. If the user's vehicle is equipped with a smart car system and has been connected to the parking lot management system, the application also provides a one-key car retrieval function. After the user confirms, the vehicle can automatically drive to the user's specified pickup point, such as the elevator hall or the exit. During the entire car search process, the system pushes real-time updates, including fine-tuning of the vehicle's location, reminders of exit congestion status, and information about promotional activities in the parking lot, helping users make efficient decisions. When the user arrives at the vehicle location and starts the vehicle, the system detects the vehicle movement through radar or camera, automatically updates the parking space status to idle, and records the departure time, completing the reverse car search process.

[0062] It can be understood that through real-time path planning and multi-sensor data fusion, the accuracy and efficiency of the reverse car search process are improved, and the user's car search time is reduced; through intelligent push of real-time congestion information and personalized navigation guidance, user experience and parking lot flow optimization are improved; through the interface of the car system, the automatic car retrieval function is realized, and the intelligence and convenience of parking services are improved.

[0063] The second aspect of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the functions of the underground parking management system based on multi-sensor fusion according to any one of the embodiments of the first aspect when executing the computer program.

[0064] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the underground parking management system based on multi-sensor fusion, and connects various parts of the underground parking management system based on multi-sensor fusion through various interfaces and lines.

[0065] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the underground parking management system based on multi-sensor fusion by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required for a function (such as a sound playing function, an image playing function, etc.), etc. The data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0066] The third aspect of the present application also provides a storage medium, wherein the storage medium stores an underground parking management program based on multi-sensor fusion, and the underground parking management program based on multi-sensor fusion implements the functions of the underground parking management system based on multi-sensor fusion according to any one of the embodiments of the first aspect when executed by a processor.

[0067] Compared with the prior art, the present application has at least the following beneficial effects: The underground parking lot management system and computer equipment based on multi-sensor fusion provided by the present application can generate optimal parking space recommendations and initial paths based on weighted scores when receiving requests, thereby improving the efficiency of initial guidance and the parking experience of users; can continuously obtain multi-source sensor data after a vehicle enters the parking lot, thereby realizing full-process monitoring and improving the comprehensiveness and real-time performance of the perception of the parking lot environment and the parking space state; can fuse multi-source detection data through a preset fusion algorithm and determine the recommended parking space state in real time, thereby improving the accuracy and reliability of the determination of the parking space state; and can immediately re-execute the recommendation algorithm and update the path when determining that the parking space is unavailable, thereby improving the dynamic adjustment capability and overall service robustness of the system in the presence of abnormal conditions.

[0068] Further, the application also realizes vehicle positioning through a Bluetooth beacon network, improves positioning reliability and coverage range; improves the accuracy of target existence and motion judgment through multi-dimensional sensing of parking space state by a radar sensor; improves the accuracy of parking space and vehicle identity recognition through camera and convolutional neural network model identification of visual information; improves the stability and anti-interference ability of positioning data through Kalman filter algorithm processing of Bluetooth signal strength data; improves the accuracy and reliability of vehicle position information through positioning calculation model calculation of vehicle coordinates; improves the uniformity and fusion of information expression through conversion of multi-source data into basic probability distribution; improves the accuracy and anti-interference ability of state judgment decision through Dempster synthesis rule fusion of multi-source evidence; improves the definiteness and operability of state judgment through setting of judgment threshold comparison of comprehensive trust degree; improves the quantitative rationality of distance factor in state discrimination through generation of first probability distribution by vehicle positioning data; improves the discrimination granularity of target existence and motion characteristics through generation of second probability distribution by parking space state sensing data; improves the contribution of image evidence in fusion decision through generation of third probability distribution by visual recognition data; improves the fault tolerance of fusion results to contradictory information through calculation of conflict coefficient between evidence and normalization processing; improves the rationality and accuracy of comprehensive trust degree calculation through synthesis of evidence supporting the same state and normalization; improves the clarity and decision efficiency of state judgment output through extraction of explicit probability value from fused probability distribution; improves the rationality and user satisfaction of parking space recommendation through comprehensive consideration of destination distance, path length, estimated time consumption and user preference for weighted scoring; improves the fine degree of spatial utilization rate perception through division of parking lot into grid units to calculate real-time heat value; improves data stability and trend prediction ability through fusion of historical heat value to generate a comprehensive heat map; improves the monitoring efficiency of parking lot operation state through visual presentation of regional heat distribution; improves the fine and energy efficiency of lighting control through differential configuration of lighting parameters according to heat map activity level differences; improves driving safety and guidance through lighting of upstream area lamps in advance by the vehicle triggering distance; improves visual comfort and reduces energy consumption through smooth adjustment of brightness and maintenance time design.

[0069] In summary, the underground parking lot management system and computer equipment based on multi-sensor fusion proposed by the application solve the technical problem that the prior art cannot realize dynamic guidance adjustment when the parking space state is abnormal due to inaccurate perception and delayed decision.

[0070] The above-described embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but should not be understood as a limitation on the patent scope of the present application. It should be noted that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and equivalent structural transformations made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields are all included in the patent protection scope of the present application. Therefore, the patent protection scope of the present application should be subject to the appended claims.

[0071] It should be noted that, in the present application, the embodiments implemented on the side of the underground parking management system based on multi-sensor fusion can be mutually referenced with the embodiments implemented on the side of the underground parking management method based on multi-sensor fusion, and the present application will not be described one by one.

[0072] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or direct or indirect application in other related technical fields are all included in the patent protection scope of the present application.

Claims

1. A multi-sensor fusion based underground parking management system, characterized in that, The application relates to a parking space recommendation and navigation system, which comprises the following modules: a parking space recommendation and navigation module, which is used for receiving a user parking request, generating a parking space recommendation result by a weighting score algorithm based on preset evaluation indexes, and generating an initial navigation path from a current position of the user to a recommended parking space; a data acquisition module, which is used for continuously acquiring multi-source detection data from multiple types of sensors in response to detection of entry of a user vehicle into a preset monitoring area of a parking lot, wherein the multi-source detection data comprises vehicle positioning data, parking space state sensing data and visual recognition data; a fusion determination module, which is used for determining a state of the recommended parking space in real time according to a preset parking space data fusion algorithm and the multi-source detection data; a dynamic scheduling module, which is used for re-executing the weighting score algorithm based on real-time multi-source detection data and preset evaluation indexes, generating a new parking space recommendation result and updating a navigation path in response to the state of the recommended parking space changing to be unavailable.

2. The multi-sensor fusion based underground parking management system as claimed in claim 1, wherein, The data acquisition module comprises: a data triggering unit, which is used for triggering a data acquisition process when it is detected that a user vehicle enters the preset monitoring area; a positioning data acquisition unit, which is used for acquiring vehicle positioning data reported by a user terminal, wherein the user terminal receives signals of multiple Bluetooth beacons in a parking lot and calculates the vehicle positioning data based on a positioning algorithm; a radar data acquisition unit, which is used for acquiring parking space state sensing data by a radar sensor arranged above a parking space, wherein the parking space state sensing data comprises existence state, motion state and contour feature information of a target object on the parking space; a visual data acquisition unit, which is used for acquiring visual data by a camera module arranged in the parking lot, performing image recognition processing on the visual data by adopting a target detection model based on a convolutional neural network, and generating visual recognition data comprising parking space number and license plate number recognition results.

3. The multi-sensor fusion based underground parking management system as claimed in claim 2, wherein, The positioning data acquisition unit comprises: a signal receiving subunit, which is used for controlling the user terminal to receive signal strength data from multiple Bluetooth beacons; a filtering processing subunit, which is used for processing the signal strength data by adopting a Kalman filtering algorithm to filter out noise and improve positioning stability; a positioning solution subunit, which is used for calculating vehicle position coordinates by a positioning solution model based on the processed signal strength data, and generating the vehicle positioning data.

4. The multi-sensor fusion based underground parking management system as claimed in claim 3, wherein, The fusion determination module comprises: a probability distribution generation unit, which is used for generating multiple basic probability distributions corresponding to parking space occupation states based on the vehicle positioning data, the parking space state sensing data and the visual recognition data; an evidence fusion unit, which is used for performing fusion calculation on the multiple basic probability distributions based on a Dempster combination rule to obtain a comprehensive trust degree of the recommended parking space being occupied; a state determination unit, which is used for comparing the comprehensive trust degree with a preset determination threshold, and determining the state of the recommended parking space according to a comparison result.

5. The multi-sensor fusion based underground parking management system as claimed in claim 4, wherein, The probability distribution generation unit comprises: a first distribution subunit, which is used for calculating a distance relationship between a user vehicle and the recommended parking space based on the vehicle positioning data, and generating a first basic probability distribution; a second allocation subunit configured to generate a second basic probability distribution based on a target presence state and a motion feature in the parking space state perception data; a third allocation subunit configured to generate a third basic probability distribution based on a parking space number and a license plate number recognition result in the visual recognition data.

6. The multi-sensor fusion based underground parking management system as claimed in claim 5, wherein, The evidence fusion unit includes: a conflict calculation subunit configured to calculate a conflict coefficient between the first basic probability distribution, the second basic probability distribution, and the third basic probability distribution; a synthesized evidence generation subunit configured to synthesize all evidence in the plurality of basic probability distributions supporting a "recommended parking space is occupied" state to obtain a preliminary synthesized evidence, and calculate a normalization factor using the conflict coefficient; a comprehensive probability distribution subunit configured to divide the preliminary synthesized evidence by the normalization factor to obtain a fused comprehensive basic probability distribution; a trust degree extraction subunit configured to extract a probability value corresponding to the "recommended parking space is occupied" state from the comprehensive basic probability distribution as the comprehensive trust degree.

7. The multi-sensor fusion based underground parking management system as claimed in claim 1, wherein, The parking recommendation and navigation module includes: a request receiving unit configured to receive a parking request submitted by a user through a mobile terminal, the request including destination information and a parking space preference type; a parking space screening unit configured to obtain real-time parking space state data of a parking lot and screen a set of available parking spaces; a score calculation unit configured to calculate a weighted score of each parking space in the set of available parking spaces based on a distance to the destination, a path planning distance, an estimated time consumption, and a parking space preference matching degree; a recommendation generation unit configured to sort the parking spaces according to the weighted score results and recommend a parking space with the highest score to the user; a path planning unit configured to generate an optimal navigation path based on a current location of the user and location information of the recommended parking space.

8. The multi-sensor fusion based underground parking management system according to any one of claims 1 to 7, wherein, The system further includes a heat map generation module configured to generate a parking lot area heat map based on state determination results of all parking spaces, the heat map generation module including: a grid division unit configured to divide the parking lot area into a plurality of uniform grid units; a real-time heat value calculation unit configured to calculate a current heat value based on real-time state data of parking spaces in each grid unit; a heat value fusion unit configured to perform weighted fusion of the current heat value and a historical heat value to generate a comprehensive heat value of each grid unit; a visualization generation unit configured to generate a visual heat map based on the comprehensive heat values of all grid units.

9. The multi-sensor fusion based underground parking management system as claimed in claim 8, wherein, The system further includes an intelligent lighting control module configured to dynamically adjust a lighting control strategy of each region according to an activity level of the region in the parking lot area heat map, the intelligent lighting control module including: an activity level determination unit configured to determine an activity level of each grid unit in the parking lot area heat map based on a comprehensive heat value of the grid unit; a parameter configuration unit configured to query a predefined lighting control parameter mapping table according to the activity level and configure a corresponding lighting control parameter group for each grid unit, the lighting control parameter group including a basic brightness value, a vehicle triggering distance, a brightness increase time, and a brightness maintenance time. a brightness boost control unit for controlling the luminaries of the grid cell and adjacent upstream grid cells to smoothly adjust from the base brightness value to 100% illumination brightness value according to the brightness boost time when a vehicle trigger distance of a target entering the grid cell is detected; a brightness drop control unit for maintaining illumination according to the brightness maintenance time after the target is detected to leave, and then smoothly reducing the brightness to the base brightness value according to a pre-set brightness drop curve.

10. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor, when executing the computer program, implements the function of the underground parking lot management system according to any one of claims 1 to 9.

Images