CN-122024518-A - Parking management and control method, system and device based on cognitive arbitration

CN122024518ACN 122024518 ACN122024518 ACN 122024518ACN-122024518-A

Abstract

The invention belongs to the technical field of parking management, and particularly relates to a parking management and control method, system and device based on cognitive arbitration, which comprise the steps of obtaining a current observation vehicle image and calculating a cognitive uncertainty score; the method comprises the steps of performing reachable judgment, screening candidate vehicles to construct a dispute candidate set, calculating a memory strength value and a competition strength index, combining a cognitive uncertainty score, weighting and summing to obtain an arbitration priority index, extracting a stable common area and an exclusive difference area for each candidate vehicle in the current observed vehicle and the dispute candidate set, mapping the stable common area and the exclusive difference area to be masks, calculating a common path drift amount and a differential path drift amount through double-path recoding, calculating a candidate resolution score in combination with the memory strength value, selecting the highest candidate vehicle as a final candidate vehicle, outputting a same vehicle, different vehicle or pending result, and executing a pass, intercept or virtual blocking instruction according to the result. The invention improves the accuracy of vehicle identity judgment in complex parking scenes.

Inventors

Wei Yingchen
FENG XIAO
WANG RANRAN
LI SHANSHAN
GU XIUJUN
YUAN MINGXU

Assignees

泰安市东信智联信息科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. The parking management and control method based on cognitive arbitration is characterized by comprising the following steps of: S1, acquiring a current observation vehicle image, performing multiple random forward reasoning on the current observation vehicle image to obtain a feature vector set, calculating a feature vector mean value, and calculating a cognitive uncertainty score of the current observation vehicle based on a difference value between each feature vector and the feature vector mean value; S2, according to the observation time, the observation position and the parking lot topological relation of the current observation vehicle, carrying out accessibility judgment on vehicles in the historical vehicle set, wherein the vehicles meeting the accessibility judgment are used as candidate vehicles, sorting the candidate vehicles according to the similarity of rough matching from high to low, selecting the candidate vehicles with the highest similarity and the target number to form a dispute candidate set, respectively calculating the memory strength value of each candidate vehicle in the dispute candidate set, and calculating the competition strength index according to the matching score between the current observation vehicle and the candidate vehicles with the target number; obtaining an arbitration priority index based on the standard deviation weighted sum of the cognitive uncertainty score, the competition strength index and the memory strength value, and executing S3 when the arbitration priority index exceeds a priority index threshold; S3, respectively constructing a dispute vehicle pair for each candidate vehicle in the current observation vehicle and the dispute candidate set, extracting a stable common area and an exclusive difference area of each dispute vehicle, respectively mapping the stable common area and the exclusive difference area into a common mask and a difference mask, respectively carrying out double-path recoding on feature vectors of the current observation vehicle and each candidate vehicle according to the common mask and the difference mask, calculating a common path drift amount and a difference path drift amount corresponding to each candidate vehicle, calculating a candidate resolution score by combining a memory intensity value of each candidate vehicle, selecting a candidate vehicle with the highest candidate resolution score as a final candidate vehicle, and obtaining a same vehicle result, a different vehicle result or a pending result according to the common path drift amount, the difference path drift amount and the candidate resolution score of the final candidate vehicle; S4, if a same car result is obtained, a release instruction is sent out; If the result of the different vehicles is obtained, an interception instruction is sent out; and if the undetermined result is obtained, a virtual blocking instruction is sent out.
2. The cognitive arbitration-based parking management and control method according to claim 1, wherein in S2, according to the observation time, the observation position and the parking lot topological relation of the current observed vehicle, the accessibility of the vehicles in the historical vehicle set is determined, specifically: Vehicles in a historical vehicle collection Is the reachable determination result of (1) Determined according to the following formula: , Wherein, the For vehicles in a historical vehicle collection Is used for judging whether the judging result is the same as the judging result, For vehicles in a historical vehicle collection Parking lot topology path length from last observed location to current observed location, For vehicles in a historical vehicle collection The time interval from the last observation time to the current observation time, The reasonable maximum passing speed of the vehicle is set for the parking lot; Only when And when the corresponding historical vehicles are included in the candidate set to serve as candidate vehicles.
3. The parking management and control method based on cognitive arbitration according to claim 1, wherein the memory strength value is calculated in S2, specifically: , Wherein, the In order to memorize the intensity value of the light, And Is a weight coefficient of the memory intensity value, and , Is a time decay coefficient; For candidate vehicles Is a time interval of (2); For candidate vehicles Is used for the track continuity scoring of the (c), For candidate vehicles Is a historical appearance stability score for (1); Candidate vehicle Trajectory continuity scoring of (2) Calculated according to the following formula: , Wherein, the For candidate vehicles The number of effective track segments meeting topological adjacency and time continuous constraint in a preset time window; For candidate vehicles The number of all track segments in the preset time window; Candidate vehicle Historical appearance stability score of (2) Calculated according to the following formula: , Wherein, the For candidate vehicles Historical observation times of (a); For candidate vehicles First, the Observing corresponding feature vectors by secondary histories; For candidate vehicles Is a historical average feature vector of (1); Is a feature distance function.
4. The parking management and control method based on cognitive arbitration according to claim 1, wherein in S3, a same-vehicle result, a different-vehicle result or a pending result is obtained according to a common path drift amount, a difference path drift amount and a candidate resolution score of a final candidate vehicle, specifically: when the final candidate vehicle meets the common path drift amount Common path determination threshold value, differential path drift amount Differential path decision threshold, and candidate resolution score Judging as a co-vehicle result when the candidate resolution threshold value; When all the candidate vehicles do not meet the same-vehicle judging condition, and at least one candidate vehicle meets the differential path drift amount When the difference path judges the threshold value, judging that the vehicle is different; and when the highest value of the candidate resolution score is smaller than the candidate resolution threshold, determining as a pending result.
5. The parking management and control method based on cognitive arbitration according to claim 1, wherein in S3, a common path drift amount and a differential path drift amount corresponding to each candidate vehicle are calculated, and a candidate resolution score is calculated by combining the memory strength values of each candidate vehicle, specifically: Candidate resolution score Calculated according to the following formula: , Wherein, the And The weighting coefficients for the candidate resolution scores, Common path drift amount for the memory intensity value of the ith candidate vehicle And differential path drift amount Calculated according to the following formula respectively: , , Wherein, the For the feature vector of the currently observed vehicle, Is the first The feature vectors of the individual candidate vehicles are, And The current observed vehicle feature vector and the first observed vehicle feature vector are enhanced by the commonality mask respectively The feature vectors of the respective candidate vehicles are, And The current observed vehicle feature vector and the i-th candidate vehicle feature vector after the difference mask enhancement, Is a feature distance function.
6. The parking management and control method based on cognitive arbitration according to claim 1, wherein the step of calculating a cognitive uncertainty score of a current observed vehicle in S1 is specifically as follows: let the current observation vehicle image warp Sub-random forward reasoning to obtain feature vector set The feature vector average value is Cognitive uncertainty score Calculated according to the following formula: , Wherein, the For the number of random forward inferences, Is the first The feature vector derived by the sub-random forward reasoning, Is that Sub-random forward reasoning of the mean vector of the corresponding feature vectors, Representing the multiplication by element, And 2 norms are represented, and U is a cognitive uncertainty score.
7. The parking management and control method based on cognitive arbitration according to claim 1, wherein in S2, a competition strength index is calculated according to a matching score between a current observed vehicle and a target number of candidate vehicles, specifically: , For the highest matching score between the current observed vehicle and the target number of candidate vehicles, For a match score between the current observed vehicle and the candidate vehicle i, To prevent the denominator from being a constant of zero, Q is the competitive strength index.
8. The cognitive arbitration-based parking control method as claimed in claim 1, wherein if a pending result is obtained in S4, calculating a congestion index of a current exit; When congestion index When the congestion threshold value is reached, the outlet of the parking lot is judged to be in a congestion state, when congestion index And when the congestion threshold value is reached, judging that the exit of the parking lot is in a non-congestion state, When the vehicle is in a non-congestion state, an interception instruction is sent out, physical interception is executed, and whether the vehicle is the same vehicle or different is checked; and executing the virtual blocking when the vehicle is in the congestion state, and executing physical release of the vehicle with the determined result in the virtual blocking process.
9. A cognitive arbitration based parking management system for implementing a cognitive arbitration based parking management method as set forth in any one of claims 1-8, comprising: The cognitive uncertainty score calculation module is used for acquiring a current observation vehicle image, executing random forward reasoning for a plurality of times on the current observation vehicle image to obtain a feature vector set, calculating a feature vector mean value, and calculating a cognitive uncertainty score of the current observation vehicle based on the difference value between each feature vector and the feature vector mean value; The competition strength index calculation module is used for carrying out accessibility judgment on vehicles in the historical vehicle set according to the observation time, the observation position and the parking lot topological relation of the current observation vehicle, and ordering the candidate vehicles according to the similarity of rough matching from high to low, selecting the candidate vehicles with the highest similarity and the target number to form a competition candidate set, respectively calculating the memory strength value of each candidate vehicle in the competition candidate set, and calculating the competition strength index according to the matching score between the current observation vehicle and the candidate vehicles with the target number; Obtaining an arbitration priority index based on the standard deviation weighted sum of the cognitive uncertainty score, the competition strength index and the memory strength value, and executing result judgment when the arbitration priority index exceeds a priority index threshold; The result judging module respectively constructs a dispute vehicle pair for each candidate vehicle in the current observation vehicle and the dispute candidate set, extracts a stable common area and an exclusive difference area of each dispute vehicle and respectively maps the stable common area and the exclusive difference area to a common mask and a difference mask, respectively carries out double-path recoding on the feature vectors of the current observation vehicle and each candidate vehicle according to the common mask and the difference mask, calculates a common path drift amount and a difference path drift amount corresponding to each candidate vehicle, calculates a candidate resolution score by combining the memory intensity value of each candidate vehicle, and selects the candidate vehicle with the highest candidate resolution score as a final candidate vehicle, and obtains a same vehicle result, a different vehicle result or a pending result according to the common path drift amount, the difference path drift amount and the candidate resolution score of the final candidate vehicle; The processing module is used for sending a release instruction if a same car result is obtained, sending an interception instruction if a different car result is obtained, and sending a virtual blocking instruction if a to-be-determined result is obtained.
10. A cognitive arbitration based parking management apparatus comprising a processor and a memory, wherein the processor implements a cognitive arbitration based parking management method as claimed in any one of claims 1 to 8 when executing a computer program stored in the memory.

Description

Parking management and control method, system and device based on cognitive arbitration Technical Field The invention belongs to the technical field of parking management, and particularly relates to a cognitive arbitration-based parking management and control method, system and device. Background The existing parking management and control system generally realizes vehicle entrance, exit and billing management based on license plate recognition or vehicle appearance comparison. However, in the actual parking lot operation, the following technical problems are that under the complex scenes of shielding, pollution, backlight, heavy rain, vehicle following, card flushing, card sleeving, card changing and the like, the high-confidence erroneous judgment is easy to occur by simply relying on license plates or conventional visual similarity comparison. Particularly when a large number of vehicles with the same vehicle type and the same color exist in the parking lot, the identities of different vehicles are difficult to accurately distinguish only by virtue of the appearance characteristics. And the existing scheme is more in the result complement judging level, and lacks hierarchical screening, unique candidate resolution and interpretable arbitration mechanism aiming at multiple candidate competition scenes. When a plurality of historical vehicles are highly similar to the current observed vehicle, the existing scheme can not converge to a unique identity conclusion from a plurality of candidate identities, so that a multi-candidate dilemma is caused. Meanwhile, cloud arbitration lacks interpretability, semantic evidence cannot be provided, and a large number of low-value samples are uploaded to cause network overhead and calculation delay. Disclosure of Invention The invention aims to provide a parking management and control method based on cognitive arbitration, which comprises the following steps of: S1, acquiring a current observation vehicle image, performing multiple random forward reasoning on the current observation vehicle image to obtain a feature vector set, calculating a feature vector mean value, and calculating a cognitive uncertainty score of the current observation vehicle based on the difference value of each feature vector and the feature vector mean value. S2, according to the observation time, the observation position and the parking lot topological relation of the current observation vehicle, carrying out accessibility judgment on vehicles in the historical vehicle set, taking the vehicles meeting the accessibility judgment as candidate vehicles, sorting the candidate vehicles according to the similarity of rough matching from high to low, selecting the candidate vehicles with the highest similarity and the target number to form a dispute candidate set, respectively calculating the memory strength value of each candidate vehicle in the dispute candidate set, and calculating the competition strength index according to the matching scores between the current observation vehicle and the candidate vehicles with the target number. And (3) obtaining an arbitration priority index based on the standard deviation weighted sum of the cognitive uncertainty score, the competition strength index and the memory strength value, and executing S3 when the arbitration priority index exceeds a priority index threshold. S3, respectively constructing a dispute vehicle pair for each candidate vehicle in the current observation vehicle and the dispute candidate set, extracting a stable common area and an exclusive difference area of each dispute vehicle, respectively mapping the stable common area and the exclusive difference area into a common mask and a difference mask, respectively carrying out double-path recoding on feature vectors of the current observation vehicle and each candidate vehicle according to the common mask and the difference mask, calculating a common path drift amount and a difference path drift amount corresponding to each candidate vehicle, calculating a candidate resolution score by combining a memory intensity value of each candidate vehicle, selecting the candidate vehicle with the highest candidate resolution score as a final candidate vehicle, and obtaining a same vehicle result, a different vehicle result or a pending result according to the common path drift amount, the difference path drift amount and the candidate resolution score of the final candidate vehicle. S4, if a same car result is obtained, a release instruction is sent out; If the result of the different vehicles is obtained, an interception instruction is sent out; and if the undetermined result is obtained, a virtual blocking instruction is sent out. In S2, according to the observation time, the observation position and the topological relation of the parking lot of the current observation vehicle, the reachability judgment is carried out on the vehicles in the historical vehicle set, specifically: Vehicles in a histor