CN-121982789-A - Non-inductive passing charging method for license plate recognition, mobile phone signaling and APP verification

CN121982789ACN 121982789 ACN121982789 ACN 121982789ACN-121982789-A

Abstract

The invention relates to the technical field of intelligent traffic charging, and discloses a non-inductive traffic charging method for license plate recognition, mobile phone signaling and APP verification, which comprises the steps of recognizing formation vehicles and acquiring signaling data; extracting signaling characteristics, acquiring a three-dimensional road network and base station coverage model, calculating signaling vertical height probability distribution, generating an overhead layer signaling set and a ground layer signaling set according to hierarchical grouping signaling, constructing a hierarchical space-time relevance scoring matrix, executing cross-hierarchical joint optimal matching by using a hierarchical constraint extended Hungary algorithm to output a vehicle-signaling-hierarchical triplet pairing result, restoring a three-dimensional path based on a hidden Markov model, calculating hierarchical segmentation cost and generating a fee deduction request. The method solves the problems of signaling attribution identification and three-dimensional path restoration when the formation vehicles pass in the complex interchange area, and realizes accurate calculation of road passing cost of different levels.

Inventors

ZHAO YANDONG
JIN GUANG
Wen Keyao
LIU QIUFENG
WANG TIANJUE

Assignees

智路云(辽宁)交通科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260205

Claims (10)

1. The non-inductive passing charging method for license plate recognition, mobile phone signaling and APP verification is characterized by comprising the following steps: acquiring a vehicle image sequence acquired by a charging point, identifying license plate numbers and a passing time stamp, marking continuous vehicles with passing time intervals smaller than a preset threshold as a formation group, retrieving relevant mobile phone identifications according to each license plate number in the formation group, and acquiring signaling record sets of all relevant mobile phones in a formation time window; extracting the characteristics of each signaling record in the signaling record set, obtaining the number of the base station, the signal strength value and the switching time stamp, and generating a signaling characteristic sequence; Acquiring three-dimensional road network topology data corresponding to a charging area and a three-dimensional coverage model of each base station; calculating the vertical height probability distribution of each signaling moment based on the signal intensity value recorded by each signaling and the three-dimensional coverage model of the corresponding base station; Grouping the signaling records according to the peak value height of the vertical height probability distribution to generate an overhead layer signaling set and a ground layer signaling set; Dividing formation vehicles into an overhead traffic vehicle set and a ground traffic vehicle set, and respectively constructing a first space-time correlation scoring matrix between the overhead layer signaling set and the overhead traffic vehicle set and a second space-time correlation scoring matrix between the ground layer signaling set and the ground traffic vehicle set; Performing cross-level joint optimal matching processing on the first time-space association degree scoring matrix and the second time-space association degree scoring matrix, and outputting a vehicle-signaling-level triplet pairing result set; Constructing a hidden Markov model containing level constraint based on three-dimensional road network topology data, searching an optimal path sequence in a level by utilizing a Viterbi algorithm, and generating three-dimensional complete passing paths of all vehicles; And calculating the toll of each road section according to the three-dimensional complete toll path, accumulating and generating the total toll of the vehicle, generating a toll deduction request and sending the toll deduction request to a user account system.
2. The method for non-inductive toll collection by license plate recognition, mobile phone signaling and APP verification according to claim 1, wherein the calculating of the vertical height probability distribution at each signaling time comprises: Based on a wireless signal propagation attenuation model, mapping a received signal intensity value into a distance estimation value between a transmitting source and a receiving point, and calculating posterior probability of the receiving point at different vertical heights by combining the installation height and the vertical directivity gain characteristic of a base station antenna; The posterior probability of the vertical height under the condition that the observed signal strength is equal to the product of the likelihood probability of the received signal strength at the height and the prior probability of the height, divided by the sum of the likelihood probability and the prior probability products at all candidate heights; Modeling the likelihood probability into Gaussian distribution according to a theoretical signal intensity mean value and a measurement error standard deviation calculated by a logarithmic distance path loss model; the prior probability is determined according to road distribution density at each vertical height in the toll area.
3. The method for non-inductive toll collection by license plate recognition, mobile phone signaling and APP verification according to claim 2, wherein when signal intensity observations of a plurality of base stations exist at the same signaling moment, vertical height probability distribution of each base station is subjected to weighted fusion, and the fused vertical height probability is equal to the sum of products of the vertical height probability of each base station and corresponding weight coefficients divided by the sum of weight coefficients; The weight coefficient is inversely proportional to the square of the horizontal distance of the base station to the charging point.
4. The license plate recognition and mobile phone signaling and APP verification non-inductive toll collection method of claim 1, wherein grouping signaling records according to peak height of vertical height probability distribution comprises: Extracting peak heights of vertical height probability distribution functions corresponding to all signaling records, comparing the peak heights with a level height threshold, classifying signaling records with peak heights higher than the level height threshold into an overhead layer signaling set, and classifying signaling records with peak heights lower than or equal to the level height threshold into a ground layer signaling set; when the vertical height probability distribution function of the signaling record has a plurality of peaks and the heights of the peaks are respectively positioned at different levels, marking the signaling record as a signaling to be determined and simultaneously keeping the signaling record in an overhead layer signaling set and a ground layer signaling set; The multimodal judgment standard is that probability values of two or more local maximum points in the vertical height probability distribution function are larger than a preset proportion threshold value of a global maximum probability value.
5. The method for non-inductive toll collection with license plate recognition and cell phone signaling and APP verification of claim 1, wherein constructing the first and second space-time relevance scoring matrices comprises: normalizing the time difference between the vehicle passing time stamp and the signaling switching time stamp, and normalizing the space distance between the vehicle passing position and the coverage area of the base station; for each vehicle-signaling pair in each scoring matrix, the spatio-temporal correlation score value is equal to the sum of the gaussian attenuation value of the temporal correlation weight times the normalized temporal difference and the gaussian attenuation value of the spatial correlation weight times the normalized spatial distance; the Gaussian attenuation value is equal to a natural exponential function value which is obtained by dividing the square of a negative difference value by the square of a double attenuation coefficient; the sum of the time-associated weight and the space-associated weight is equal to one.
6. The method for non-inductive toll collection by license plate recognition, mobile phone signaling and APP verification according to claim 1, wherein the cross-level joint optimal matching process comprises: Longitudinally splicing the first space-time relevance scoring matrix and the second space-time relevance scoring matrix to construct a joint scoring matrix; Setting a cross-level constraint condition in the joint scoring matrix, setting a scoring value between an overhead passing vehicle and a ground layer signaling to be minus infinity, setting a scoring value between the ground passing vehicle and the overhead layer signaling to be minus infinity, and generating a constrained joint scoring matrix; Performing a Hungary algorithm on the constrained joint scoring matrix, and searching for a vehicle-signaling one-to-one pairing scheme which maximizes the total score; And according to the original hierarchy attribution of each signaling record in the pairing scheme, attaching a hierarchy label to each vehicle-signaling pairing, and generating a vehicle-signaling-hierarchy triplet pairing result set.
7. The license plate recognition and mobile phone signaling and APP verification non-inductive passing charging method according to claim 6, wherein for signaling records marked as signaling to be determined, pairing of overhead passing vehicles or ground passing vehicles is allowed in the execution process of a Hungary algorithm, and final level attribution of the signaling to be determined is determined according to the level of paired vehicles after pairing is completed; When the number of the formation vehicles is inconsistent with the number of the signaling, performing expansion processing on the joint scoring matrix before executing the Hungary algorithm, adding a virtual signaling column or a virtual vehicle row, setting the scoring value of the virtual item to be a preset low score value, and marking the item paired with the virtual item as an unmatched state after matching is completed.
8. The license plate recognition and phone signaling and APP verification non-inductive toll collection method of claim 1 wherein said constructing a hidden markov model containing hierarchical constraints comprises: extracting a road segment set corresponding to the hierarchical label from the three-dimensional road network topology data to serve as a candidate road segment set; Constructing a hidden Markov model based on the candidate road segment set, wherein the state nodes correspond to each road segment in the candidate road segment set, the state transition probability is determined according to the topological connection relation of the adjacent road segments and the running direction of the vehicle, and the observation probability is determined according to the base station number recorded by signaling and the space coverage relation of the road segments; The state transition probability is determined in such a way that if two road sections have a direct connection relationship and the running directions are consistent, the state transition probability is set to a preset high probability value which is adjusted according to the steering angle between the adjacent road sections; the observation probability is determined in such a way that if the intersection exists between the horizontal coverage area polygon of the base station and the horizontal projection of the road section, the observation probability is set to be a value positively related to the intersection area, and if the intersection does not exist, the observation probability is set to be zero.
9. The license plate recognition and mobile phone signaling and APP verification non-inductive toll collection method according to claim 8, wherein a jump transfer mechanism is introduced in the hidden markov model, when the time interval of adjacent signaling records exceeds a preset interval threshold, the state node is allowed to jump to a road section which is not directly adjacent but has reachable topology, and the jump transfer probability is equal to the preset basic probability value multiplied by the damping factor to the power of the number of jump road sections.
10. A non-inductive toll collection system for vehicles forming a complex interchange area, for executing the license plate recognition and mobile phone signaling and APP verification non-inductive toll collection method of any one of claims 1 to 9, comprising: The formation recognition and signaling acquisition module is used for acquiring a vehicle image sequence acquired by a charging point, recognizing license plate numbers and passing time stamps, marking continuous vehicles with passing time intervals smaller than a preset threshold as a formation group, searching associated mobile phone identifiers according to the license plate numbers in the formation group, and acquiring signaling record sets of all associated mobile phones in a formation time window; the signaling feature extraction module is used for carrying out feature extraction on each signaling record in the signaling record set, obtaining the number of the base station, the signal strength value and the switching time stamp, and generating a signaling feature sequence; the three-dimensional model acquisition module is used for acquiring three-dimensional road network topology data corresponding to the charging area and a three-dimensional coverage model of each base station; The vertical height probability calculation module is used for calculating the vertical height probability distribution of each signaling moment based on the signal intensity value recorded by each signaling and the three-dimensional coverage model of the corresponding base station; The signaling layering grouping module is used for grouping signaling records according to the peak value height of the vertical height probability distribution to generate an overhead layer signaling set and a ground layer signaling set; The hierarchical scoring matrix construction module is used for dividing the formation vehicles into an overhead traffic vehicle set and a ground traffic vehicle set, and respectively constructing a first space-time correlation scoring matrix between the overhead layer signaling set and the overhead traffic vehicle set and a second space-time correlation scoring matrix between the ground layer signaling set and the ground traffic vehicle set; the cross-level matching module is used for performing cross-level joint optimal matching processing on the first time-space association degree scoring matrix and the second time-space association degree scoring matrix and outputting a vehicle-signaling-level triplet pairing result set; The three-dimensional path restoration module is used for constructing a hidden Markov model containing level constraint based on three-dimensional road network topology data, searching an optimal path sequence in a level by utilizing a Viterbi algorithm, and generating three-dimensional complete passing paths of all vehicles; The fee calculation and deduction module is used for calculating the toll of each road section according to the three-dimensional complete toll path, accumulating and generating the total toll of the vehicle, generating a deduction request and sending the deduction request to the user account system.

Description

Non-inductive passing charging method for license plate recognition, mobile phone signaling and APP verification Technical Field The invention relates to the technical field of intelligent traffic charging, in particular to a non-inductive traffic charging method for license plate recognition, mobile phone signaling and APP verification. Background In the complex interchange area of the city, the overhead expressway and the ground road are highly overlapped on horizontal projection, but the actual driving path and the charging standard are different. When multiple vehicle formation such as logistics vehicle formation, tourist bus formation and the like pass through a complex interchange area, the existing non-inductive passing charging system adopts a path reduction technology based on license plate recognition and mobile phone signaling to conduct charging treatment. The prior art has the following defects that the passing time interval of formation vehicles is extremely short, the space distance is only tens of meters, the signaling data of a plurality of associated mobile phones are highly overlapped in the space-time dimension, the same base station covers the overhead layer and the ground layer road at the same time, and the conventional two-dimensional road network topology cannot express the vertical hierarchical relationship. The problem is that two-dimensional formation matching cannot distinguish the attribution of vehicles on paths of different levels, and a single-vehicle three-dimensional path restoration method cannot accurately distinguish signaling sources of each vehicle in a formation scene, so that the technical problems of path restoration errors and cost calculation errors are finally caused. Disclosure of Invention The invention provides a non-inductive passing charging method for license plate recognition, mobile phone signaling and APP verification, which solves the technical problems of inaccurate vehicle passing path recognition and low charging efficiency of formation vehicles caused by the aliasing of an overhead road layer and a ground road layer in a complex interchange area in the related technology. The invention discloses a non-inductive passing charging method for license plate recognition, mobile phone signaling and APP verification, which comprises the following steps of acquiring a vehicle image sequence acquired by a charging point, recognizing license plate numbers and passing time stamps, marking continuous vehicles with passing time intervals smaller than a preset threshold value as a formation group, searching relevant mobile phone identifiers according to each license plate number in the formation group, and acquiring signaling record sets of all relevant mobile phones in a formation time window; the method comprises the steps of extracting characteristics of each signaling record in a signaling record set, obtaining a base station number, a signal intensity value and a switching time stamp, generating a signaling characteristic sequence, obtaining three-dimensional road network topological data corresponding to a charging area and a three-dimensional coverage model of each base station, calculating vertical height probability distribution of each signaling moment based on the signal intensity value of each signaling record and the three-dimensional coverage model of the corresponding base station, grouping the signaling records according to peak heights of the vertical height probability distribution, generating an overhead layer signaling set and a ground layer signaling set, dividing formation vehicles into the overhead communication vehicle set and the ground communication vehicle set, respectively constructing a first time-space association degree scoring matrix between the overhead layer signaling set and the overhead communication vehicle set and a second time-space association degree scoring matrix between the ground layer signaling set and the ground communication vehicle set, performing cross-level joint optimal matching treatment on the first time-space association degree scoring matrix and the second time-space association degree scoring matrix, outputting a hidden Markov model-signaling-level ternary combination pairing result set, constructing a hidden Markov model containing level constraint based on the three-dimensional road network data, utilizing a Viterbi sequence algorithm, and calculating the passing cost of each road section according to the three-dimensional complete passing path, accumulating to generate the total passing cost of the vehicles, generating a fee deduction request and sending the fee deduction request to a user account system. The method for calculating the vertical height probability distribution of each signaling moment comprises the steps of mapping a received signal intensity value into a distance estimation value between a transmitting source and a receiving point based on a wireless signal propagation attenuation model,