An analysis method based on illegal boarding and disembarking of passengers on passenger vehicles

Abstract

The present invention provides an analysis method based on illegal boarding and disembarking of passengers on passenger vehicles, comprising the following steps: step S01: obtaining illegal boarding and disembarking characteristic values ​​by screening passenger monitoring data for vehicle violations; step S02: performing characteristic analysis based on the illegal boarding and disembarking characteristic values, and constructing a vehicle violation model; step S03: generating output value similarity by performing model comparison calculation on the vehicle violation model; step S04: performing deviation identification based on the output value similarity, and obtaining information of vehicles to be inspected; step S05: performing violation detection on the vehicles to be inspected based on the vehicle information to be inspected, and obtaining vehicle violation information; by establishing a vehicle violation model, the analysis efficiency of passenger vehicle violation behaviors is improved, and the output value similarity is calculated, the accuracy of passenger vehicle violation analysis is enhanced, invalid detection is reduced, and violation detection time is saved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an analysis method based on illegal boarding and disembarking of passengers on passenger vehicles. Background Art

[0002] At present, due to the uneven quality of drivers and the frequent occurrence of illegal boarding and disembarking, the order of industry development has been affected. In addition, passengers who get off the bus arbitrarily cross the highway, which is prone to traffic accidents. Buses stop on the lanes to board and disembark illegally. Therefore, it is impossible to control buses whose license plate numbers and vehicle characteristics are unclear. In addition, due to the influence of the environment, there are more problems of irregular driving, and passengers are not picked up and disembarked according to regulations during driving, which leads to unnecessary spread of safety risks and unnecessary losses. In addition to the current ordinary roadside cameras, the existing vehicle supervision also includes "Long-distance Vehicle Based on Checkpoint Pass Information Judgment" with application number "201410692387.1". In the article "Method for detecting traffic violations of long-distance passenger vehicles", it is mentioned that highway checkpoints are used to perform spatial overlap comparison of long-distance passenger vehicles between 2 a.m. and 5 a.m., and whether there are violations is determined based on vehicle attributes; however, this technical solution is limited to being suitable for use only in a specific time period, and only reference analysis is performed on the vehicles, and it is impossible to compare the passengers getting on and off the vehicles; the present invention proposes to build a multi-algorithm capability based on AI artificial intelligence technology, analyze the data of passenger vehicles, and then perform frame extraction and algorithm analysis on real-time video to dynamically analyze the flow of people near the passenger vehicles, thereby realizing the behavior judgment of getting on and off the passengers, and through the continuous collection of data and continuous analysis of human behavior, it is possible to quickly and accurately discover violations. Summary of the invention

[0003] The present invention provides an analysis method based on illegal boarding and disembarking of passengers on passenger vehicles, so as to solve the situation.

[0004] The present invention provides an analysis method based on illegal boarding and disembarking of passengers on passenger vehicles, comprising the following steps:

[0005] Step S01: Screening the passenger transport monitoring data for vehicle violations to obtain the illegal boarding and disembarking feature values;

[0006] Step S02: Perform feature analysis based on the illegal boarding and disembarking feature values ​​to construct a vehicle violation model;

[0007] Step S03: generating output value similarity by performing model comparison calculation on the vehicle violation model;

[0008] Step S04: performing deviation identification according to the output value similarity to obtain the information of the vehicle to be inspected;

[0009] Step S05: Perform violation detection on the vehicle to be inspected according to the information of the vehicle to be inspected, and obtain the vehicle violation information

[0010] As an embodiment of the present technical solution, step S01 includes:

[0011] By classifying the passenger monitoring data according to the data function, the monitoring data category is obtained; among them,

[0012] The passenger monitoring data includes: checkpoint data and video structured data;

[0013] The monitoring data categories include: digital category, speed category, time category, and behavior category;

[0014] By performing category feature extraction on each detection data category, a category feature value of each category is generated, and passenger boarding and disembarking analysis is performed on the category feature value to obtain a feature value of illegal passenger boarding and disembarking.

[0015] As an embodiment of the present technical solution, step S02 includes:

[0016] By modeling and screening the characteristic values ​​of illegal boarding and disembarking, the characteristic information of the illegal model is obtained;

[0017] The violation model feature information includes: location information, node information, edge information, and interval information;

[0018] According to the violation model feature information, a subject model is constructed to generate a vehicle violation model.

[0019] As an embodiment of the present technical solution, the model comparison calculation includes the following steps:

[0020] Step 1: Obtain normalized data by normalizing the vehicle violation model data; wherein,

[0021] The normalization process maps the conversion result to a preset interval by normalizing the deviation of the vehicle violation model data; wherein,

[0022] The deviation standardization includes: linearly transforming the vehicle violation model data through a preset conversion function;

[0023] Step 2: Transmit the normalized data to the twin network for mapping analysis and processing to generate twin feature space data; wherein,

[0024] The twin network includes: extracting dual input data of normalized data, inputting it into the neural network, and performing mapping processing; wherein,

[0025] The dual input data includes: previous frame data and next frame data;

[0026] Step 3: perform output calculation on the twin feature space data, obtain the output value similarity, and make a judgment; wherein,

[0027] When the output value similarity is less than or equal to a preset threshold, it is a normal vehicle behavior;

[0028] When the output value similarity is greater than a preset threshold, it is abnormal vehicle behavior.

[0029] As an embodiment of the present technical solution, step S04 includes:

[0030] By matching the corresponding calculation method according to the output value similarity, performing deviation calculation, obtaining the deviation value, and making a judgment; wherein,

[0031] The deviation calculation includes: comparing the output value with a preset output comparison data set according to the corresponding calculation method to obtain a comparison deviation value;

[0032] When the deviation value is greater than a preset threshold, it is considered as an excessive deviation, and the vehicle is determined to be a to-be-detected vehicle, and the information of the to-be-detected vehicle is obtained;

[0033] When the deviation value is less than or equal to a preset threshold, it is a normal deviation.

[0034] As an embodiment of the present technical solution, step S05 includes:

[0035] The first screening information is obtained by screening the key data of the vehicle information to be inspected; wherein,

[0036] The key data screening includes: mobile status screening and path screening;

[0037] Extract vehicle information from the first screening information to obtain a first vehicle to be inspected, and perform violation detection on the first vehicle to be inspected to obtain vehicle violation information; wherein,

[0038] The violation detection includes: passenger vehicle detection and passenger vehicle license plate recognition.

[0039] As an embodiment of the present technical solution, the passenger vehicle detection includes the following steps:

[0040] Step S10: Obtain information feature layers by performing feature stratification on the vehicle information to be detected; wherein,

[0041] The feature stratification includes: scale stratification, cost stratification, and training stratification;

[0042] Step S20: extracting feature information of each information feature layer respectively according to the preset multi-scale feature learning, and generating information feature nodes corresponding to the information feature layer;

[0043] Step S30: performing subspace mapping on the information feature layer and the corresponding information feature node, obtaining a mapping result, and judging the mapping result; wherein,

[0044] When the mapping result is within a preset threshold range, it is a compliant passenger vehicle;

[0045] When the mapping result is not within the preset threshold range, it is an illegal passenger vehicle, the illegal passenger vehicle information is obtained, and the preset violation response process is performed.

[0046] As an embodiment of the present technical solution, the passenger vehicle license plate recognition comprises the following steps:

[0047] Step S100: Perform convolution feature extraction through a preset detection convolutional neural network to obtain a feature map; wherein,

[0048] The convolution feature extraction includes: convolution, activation function, and building a pooling layer;

[0049] Step S200: Performing a preset feature network analysis process according to the feature graph to generate a passenger vehicle detection frame; wherein,

[0050] The feature network analysis process includes: region proposal network processing, fully connected layer processing; wherein,

[0051] The region proposal network processing includes: generating a candidate region, obtaining a candidate region node, determining the node attributes of the candidate region, performing bounding box regression processing according to the node attributes, and determining a first candidate region; wherein,

[0052] The node attributes include: foreground node and background node;

[0053] The fully connected layer processing includes: extracting comprehensive information based on the first area to be selected to generate a feature map of the area to be selected, performing frame regression calculation based on the feature map of the area to be selected to obtain a passenger vehicle detection frame;

[0054] Step S300: Performing a passenger vehicle violation detection according to the passenger vehicle detection border to obtain passenger vehicle violation information.

[0055] As an embodiment of the present technical solution, the violation detection includes:

[0056] By performing segment detection on the passenger vehicle, segment detection information is obtained; wherein the segment detection includes: time segment detection and route segment detection;

[0057] According to the segment detection information, segment similarity analysis is performed to obtain segment similarity values ​​and make judgments; wherein,

[0058] The segmented similarity analysis is performed by calculating and analyzing the similarity between each segment of information and the previous segment of information;

[0059] When the segment similarity value is within a preset threshold range, it is a normal segmentation;

[0060] When the segment similarity value is not within the preset threshold range, it is an abnormal segment and an abnormal analysis process is performed.

[0061] As an embodiment of the present technical solution, the violation detection further includes:

[0062] By detecting the portrait violation behavior of the vehicle violation information, the portrait violation data is obtained; wherein,

[0063] The portrait violation behavior detection includes: boarding behavior detection and getting off the vehicle behavior detection;

[0064] According to the portrait violation data and the preset violation database, data comparison processing is performed to obtain the violation level; wherein,

[0065] The violation levels include: minor violation, ordinary violation, and serious violation;

[0066] According to the violation level, a corresponding security warning is issued to obtain warning information; wherein,

[0067] The safety warning includes: according to the violation level, the vehicle where the violator is located is warned online to obtain warning information;

[0068] The violating vehicle performs violation correction processing according to the warning information and generates violation correction information;

[0069] The early warning information includes: warning information, violation correction information, and positioning information.

[0070] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0071] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0073] Figure 1It is a flow chart of a method for analyzing illegal boarding and disembarking of passengers on passenger vehicles according to an embodiment of the present invention;

[0074] Figure 2 It is a flow chart of model comparison calculation in a method for analyzing illegal boarding and disembarking of passengers on passenger vehicles in an embodiment of the present invention;

[0075] Figure 3 The present invention is a flowchart of a passenger vehicle detection method based on illegal boarding and disembarking of passengers on passenger vehicles in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0077] It should be noted that when a component is referred to as being "fixed to" or "disposed on" another component, it can be directly on the other component or indirectly on the other component. When a component is referred to as being "connected to" another component, it can be directly or indirectly connected to the other component.

[0078] It should be understood that the orientation or position relationship indicated by terms such as "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and "outside" are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0079] In addition, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations, and the meaning of "plurality" is two or more, unless otherwise clearly and specifically limited. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0080] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0081] The embodiment of the present invention provides a method for analyzing illegal boarding and disembarking of passengers on a passenger vehicle, comprising the following steps:

[0082] Step SO1: Obtain the characteristic value of illegal boarding and disembarking by screening the passenger monitoring data for vehicle violations;

[0083] Step S02: Perform feature analysis based on the illegal boarding and disembarking feature values ​​to construct a vehicle violation model;

[0084] Step S03: generating output value similarity by performing model comparison calculation on the vehicle violation model;

[0085] Step S04: performing deviation identification according to the output value similarity to obtain the information of the vehicle to be inspected;

[0086] Step S05: Performing violation detection on the vehicle to be inspected according to the information of the vehicle to be inspected, and obtaining the vehicle violation information;

[0087] The working principle of the above technical scheme is as follows: In the existing technical scheme, the boarding and alighting violations of passenger vehicles are analyzed by analyzing their boarding and alighting behaviors. For example, the "A control method and system for intelligent monitoring of road passenger flow" with application number "201911312032.4" detects the boarding and alighting behaviors of passengers on the vehicle to determine whether it is overloaded, locks the vehicle through monitoring at the station and on the way, registers the illegal vehicle into the warehouse, and imposes certain penalties. It is limited to analyzing the boarding and alighting behaviors of passengers on the vehicle to determine whether the passenger vehicle has any illegal behaviors. It specifically analyzes the passenger flow, and its main purpose is to improve the safe driving of the owner and the driving safety of the passengers. The above technical scheme analyzes and judges the illegal boarding and alighting behaviors of passenger vehicles from the perspective of driving and portrait behavior. First, the monitored data is screened for violations, and the feature values ​​of illegal boarding and alighting are obtained. Then, feature analysis is performed to construct a violation model, and the model is compared and calculated to calculate the output value similarity. Deviation identification is performed based on the similarity, and the information of the vehicle to be inspected is obtained. The vehicle to be inspected is detected for violations, and finally the vehicle violation information is obtained.

[0088] The beneficial effects of the above technical solution are: in actual use, when the passenger vehicle is driving normally, the passenger vehicle data is obtained and the data is screened for violations, such as parking in a prohibited parking section, running a red light, etc., and the data is uploaded. Through screening, the efficiency of analyzing the violations of the passenger vehicle is greatly improved. By establishing a vehicle violation model and calculating the similarity of the output values, the accuracy of the analysis of passenger vehicle violations is enhanced. The vehicle to be tested is determined first, and then the vehicle to be tested is tested for violations, which reduces invalid detections, improves the efficiency of violation detection, and saves violation detection time.

[0089] In one embodiment, step S01 includes:

[0090] By classifying the passenger monitoring data according to the data function, the monitoring data category is obtained; among them,

[0091] The passenger monitoring data includes: checkpoint data and video structured data;

[0092] The monitoring data categories include: digital category, speed category, time category, and behavior category;

[0093] By extracting the category features of each detection data category, generating the category feature value of each category, performing boarding and disembarking analysis on the category feature value, and obtaining the feature value of illegal boarding and disembarking;

[0094] The working principle of the above technical solution is as follows: In the existing technical solution, the passenger vehicle data is generally directly tested to determine whether there is a violation, but this process will encounter many cases where there is no violation after testing, which reduces the efficiency of violation detection. In the above technical solution, the passenger monitoring data is classified according to the data function, and the category feature is extracted for each detection data category, and the category feature value of each category is generated, and the boarding and disembarking analysis is performed to obtain the feature value of illegal boarding and disembarking;

[0095] The beneficial effects of the above technical solution are: by classifying the passenger vehicle monitoring data, the feature extraction efficiency is greatly improved, and by extracting features and generating category feature values, data judgment is made more intuitive and the data analysis speed is improved.

[0096] In one embodiment, step S02 includes:

[0097] By modeling and screening the characteristic values ​​of illegal boarding and disembarking, the characteristic information of the illegal model is obtained;

[0098] The violation model feature information includes: location information, node information, edge information, and interval information;

[0099] According to the violation model feature information, a subject model is constructed to generate a vehicle violation model;

[0100] The working principle of the above technical solution is as follows: Different from the single screening of illegal data based on data in the existing technical solution, the above technical solution uses the illegal boarding and disembarking feature values ​​to perform modeling screening and processing, obtains the feature information of the illegal model, and then builds the model to generate a vehicle violation model;

[0101] The beneficial effects of the above technical solution are: modeling and screening are carried out according to the characteristic values ​​of boarding and disembarking passengers, which improves the model's pertinence and accuracy, and enhances the efficiency of violation screening by building a model.

[0102] In one embodiment, the model comparison calculation includes the following steps:

[0103] Step 1: Obtain normalized data by normalizing the vehicle violation model data; where

[0104] The normalization process maps the conversion result to a preset interval by performing deviation standardization on the vehicle violation model data; where

[0105] The deviation standardization includes: linearly transforming the vehicle violation model data through a preset conversion function; where

[0106] Obtain a model data group {x 1 , x 2 ,..., x t}, and the conversion function y i is:

[0107]

[0108] where y i is the conversion function of the i-th data in the model data group, x i is the i-th data in the model data group, x min is the minimum data in the model data group {x 1 , x 2 ,..., x t}, x max is the maximum data in the model data group {x 1 , x 2 ,..., x t}, i is a variable, and 1 ≤ i ≤ t, t is a constant, and 1 < t, and the mapping result is in the range of [0, 1];

[0109] Step 2: Transmit the normalized data to a siamese network for mapping analysis processing to generate siamese feature space data; where

[0110] The siamese network includes: extracting double-input data of the normalized data, inputting it into a neural network, and performing mapping processing; where

[0111] The double-input data includes: front-frame data and rear-frame data;

[0112] Step 3: Obtain a mapping data group {y 1 , y 2 ,..., y t} to perform output calculation on the siamese feature space data, and calculate the output value similarity D(X, Y);

[0113]

[0114] Among them, y i is the conversion function of the ith data in the model data set, x i is the i-th data in the model data set, p is a constant, here p=2;

[0115] When D(X, Y)≤0.95, the threshold is normal vehicle behavior;

[0116] When D(X, Y)>0.95, it is abnormal vehicle behavior;

[0117] The working principle of the above technical solution is: first, the deviation of the vehicle violation model data is normalized, and the transformation result is mapped to a preset interval to obtain normalized data. The deviation standardization includes: linearly transforming the vehicle violation model data through a preset conversion function, transmitting the normalized data to the twin network for mapping analysis and processing, and generating twin feature space data. Finally, the output calculation of the twin feature space data is performed to obtain the output value similarity. When the output value similarity is less than or equal to the preset threshold, it is a normal vehicle behavior. When the output value similarity is greater than the preset threshold, it is an abnormal vehicle behavior.

[0118] The beneficial effects of the above technical solution are: by normalizing the model data and mapping the results to the preset interval, the data accuracy is improved; by generating twin feature space data, the feature specificity is improved; by calculating the similarity of the output values, it is more intuitive to judge whether the vehicle behavior is illegal, thereby improving the accuracy of judgment.

[0119] In one embodiment, step S04 includes:

[0120] By matching the corresponding calculation method according to the output value similarity, performing deviation calculation, obtaining the deviation value, and making a judgment; wherein,

[0121] The deviation calculation includes: comparing the output value with a preset output comparison data set according to the corresponding calculation method to obtain a comparison deviation value;

[0122] The deviation calculation also includes the following steps:

[0123] Step S101: Obtain output data set {α 1 , α 2 , ..., α n}, output data affects the parameters {μ 1 , μ 2 , ..., μ n}, calculate the expected value of output data δ respectively d :

[0124]

[0125] Among them, α t is the tth output data in the output data group, μ t is the influencing parameter of the t-th output data in the output data set, t is a variable, and 1≤t≤d, d is a variable, and 1≤d≤n, n is the number of output data in the output data set, and 1≤n;

[0126] Step S102: Obtaining a control data set {β 1 , β 2 , ..., β n}, the optimal parameter of the control data {ε 1 , ε 2 , ..., ε n}, the largest impact factor Calculate the expected value τ of the control data respectively d :

[0127]

[0128] Among them, β l is the lth control data in the control data group, ε l is the optimal parameter of the lth control data in the control data group, is the maximum influencing factor of the lth control data in the control data group, l is a variable, and 1≤l≤d;

[0129] Step S103: According to the output data expected value δ d and the expected value of the control data τ d , establish a set of equations and calculate the data deviation value θ d :

[0130]

[0131] Among them, μ n+1 is the parameter with the greatest impact on data deviation, α u is the u-th output data in the output data group, θ u is the data deviation value of the u-th output data in the output data group, u is a variable, and 1≤u≤n-1, μ n is the output data influencing parameter of the nth output data in the output data group;

[0132] When the deviation value is greater than a preset threshold, it is considered as an excessive deviation, and the vehicle is determined to be a to-be-detected vehicle, and the information of the to-be-detected vehicle is obtained;

[0133] When the deviation value is less than or equal to the preset threshold, it is a normal deviation;

[0134] The working principle of the above technical solution is as follows: unlike the prior art solution, the obtained vehicle data is directly judged according to the set data, while in the above technical solution, the output value is compared and calculated with the preset output comparison data group through the calculation method corresponding to the output value similarity matching, and the comparison deviation value is obtained. When the deviation value is greater than the preset threshold, it is an excessive deviation and is determined to be a vehicle to be detected, and the information of the vehicle to be detected is obtained. When the deviation value is less than or equal to the preset threshold, it is a normal deviation;

[0135] The beneficial effects of the above technical solution are: by calculating the deviation value, the speed of determining the vehicle to be detected is improved, and the error rate of detecting non-violation vehicles is reduced.

[0136] In one embodiment, step S05 includes:

[0137] The first screening information is obtained by screening the key data of the vehicle information to be inspected; wherein,

[0138] The key data screening includes: mobile status screening and path screening;

[0139] Extract vehicle information from the first screening information to obtain a first vehicle to be inspected, and perform violation detection on the first vehicle to be inspected to obtain vehicle violation information; wherein,

[0140] The violation detection includes: passenger vehicle detection and passenger vehicle license plate recognition;

[0141] The working principle of the above technical solution is as follows: first, the detection object is the determined vehicle information to be inspected, after the mobile state screening and path screening, the first screening information is determined, and then the vehicle information is extracted to obtain the first vehicle to be inspected, and the passenger vehicle detection and passenger vehicle license plate recognition are performed to obtain the vehicle violation information;

[0142] The beneficial effects of the above technical solution are: by screening key data of the vehicles to be inspected, the efficiency of vehicle violation inspection is improved, and by determining the vehicle based on the screened information, the richness of vehicle information and the accuracy of illegal vehicles are improved.

[0143] In one embodiment, the passenger vehicle detection comprises the following steps:

[0144] Step S10: Obtain information feature layers by performing feature stratification on the vehicle information to be detected; wherein,

[0145] The feature stratification includes: scale stratification, cost stratification, and training stratification;

[0146] Step S20: extracting feature information of each information feature layer respectively according to the preset multi-scale feature learning, and generating information feature nodes corresponding to the information feature layer;

[0147] Step S30: performing subspace mapping on the information feature layer and the corresponding information feature node, obtaining a mapping result, and judging the mapping result; wherein,

[0148] When the mapping result is within a preset threshold range, it is a compliant passenger vehicle;

[0149] When the mapping result is not within the preset threshold range, it is an illegal passenger vehicle, the illegal passenger vehicle information is obtained, and the preset violation response process is performed;

[0150] The working principle of the above technical solution is as follows: Different from the existing technical solution in which the vehicle is subjected to video analysis and identification and fixed-point image analysis, the above technical solution performs feature stratification on the information of the vehicle to be detected to obtain the information feature layer, including: scale stratification, cost stratification, and training stratification. Then, according to the preset multi-scale feature learning, the feature information of each information feature layer is extracted respectively, and the information feature node corresponding to the information feature layer is generated. The information feature layer and the corresponding information feature node are subjected to subspace mapping to obtain the mapping result. When the mapping result is within the preset threshold range, it is a compliant passenger vehicle; when the mapping result is not within the preset threshold range, it is an illegal passenger vehicle, the illegal passenger vehicle information is obtained, and the preset violation response process is performed;

[0151] The beneficial effects of the above technical solution are: by performing feature stratification and subdividing the passenger vehicle detection method, the accuracy and efficiency of passenger vehicle detection are improved.

[0152] In one embodiment, the passenger vehicle license plate recognition comprises the following steps:

[0153] Step S100: Perform convolution feature extraction through a preset detection convolutional neural network to obtain a feature map; wherein,

[0154] The convolution feature extraction includes: convolution, activation function, and building a pooling layer;

[0155] Step S200: Performing a preset feature network analysis process according to the feature graph to generate a passenger vehicle detection frame; wherein,

[0156] The feature network analysis process includes: region proposal network processing, fully connected layer processing; wherein,

[0157] The region proposal network processing includes: generating a candidate region, obtaining a candidate region node, determining the node attributes of the candidate region, performing bounding box regression processing according to the node attributes, and determining a first candidate region; wherein,

[0158] The node attributes include: foreground node and background node;

[0159] The fully connected layer processing includes: extracting comprehensive information based on the first area to be selected to generate a feature map of the area to be selected, performing frame regression calculation based on the feature map of the area to be selected to obtain a passenger vehicle detection frame;

[0160] Step S300: performing a passenger vehicle violation detection according to the passenger vehicle detection frame to obtain passenger vehicle violation information;

[0161] The working principle of the above technical solution is as follows: in the existing technical solution, in special scenarios, the license plate number recognition is difficult and the license plate cannot be solved. In the above technical solution, by performing convolution feature extraction on the detection convolutional neural network, a feature map is obtained, and a preset feature network analysis and processing is performed according to the feature map to generate a passenger car detection border. The feature network analysis and processing includes: regional proposal network processing and fully connected layer processing. The regional proposal network processing generates a candidate area, obtains the node of the candidate area, determines the node attributes of the candidate area, performs surrounding border regression processing, and determines the first candidate area. The fully connected layer processing extracts comprehensive information based on the first candidate area, generates a candidate area feature map, performs border regression calculation, obtains the passenger car detection border, and finally performs passenger car violation detection to obtain passenger car violation information;

[0162] The beneficial effects of the above technical solution are: by analyzing the area to be selected and determining the passenger vehicle detection border, the accuracy of passenger vehicle violation detection is improved.

[0163] In one embodiment, the violation detection includes:

[0164] By performing segment detection on the passenger vehicle, segment detection information is obtained; wherein the segment detection includes: time segment detection and route segment detection;

[0165] According to the segment detection information, segment similarity analysis is performed to obtain segment similarity values ​​and make judgments; wherein,

[0166] The segmented similarity analysis is performed by calculating and analyzing the similarity between each segment of information and the previous segment of information;

[0167] When the segment similarity value is within a preset threshold range, it is a normal segmentation;

[0168] When the segment similarity value is not within the preset threshold range, it is considered as an abnormal segment and an abnormal analysis process is performed;

[0169] The working principle of the above technical solution is: first, segment detection is performed on passenger vehicles to obtain segment detection information, and the segment detection includes: time segment detection and route segment detection. Then, each segment information is analyzed for similarity with the previous segment information to obtain a segment similarity value. When the segment similarity value is within a preset threshold range, it is a normal segment; when the segment similarity value is not within the preset threshold range, it is an abnormal segment, and abnormal analysis processing is performed;

[0170] The beneficial effects of the above technical solution are: by performing segmented detection on passenger vehicles and calculating segmented similarity values, the accuracy and efficiency of violation detection are improved. At the same time, abnormal situations can be discovered as early as possible and abnormal response processing can be carried out, thereby improving the driving safety of passenger vehicles.

[0171] In one embodiment, the violation detection further includes:

[0172] By detecting the portrait violation behavior of the vehicle violation information, the portrait violation data is obtained; wherein,

[0173] The portrait violation behavior detection includes: boarding behavior detection and getting off the vehicle behavior detection;

[0174] According to the portrait violation data and the preset violation database, data comparison processing is performed to obtain the violation level; wherein,

[0175] The violation levels include: minor violation, ordinary violation, and serious violation;

[0176] According to the violation level, a corresponding security warning is issued to obtain warning information; wherein,

[0177] The safety warning includes: according to the violation level, the vehicle where the violator is located is warned online to obtain warning information;

[0178] The violating vehicle performs violation correction processing according to the warning information and generates violation correction information;

[0179] The early warning information includes: warning information, violation correction information, and positioning information;

[0180] The working principle of the above technical solution is as follows: unlike the existing technical solution in which the vehicle is inspected and the information of the illegal vehicle is uploaded, the above technical solution performs a portrait violation behavior detection on the vehicle violation information to obtain the portrait violation data. The portrait violation behavior detection includes: boarding behavior detection and getting off behavior detection. According to the preset violation database, data comparison processing is performed to obtain the violation level, including: minor violation, ordinary violation, and serious violation. According to the violation level, a corresponding safety warning is issued to obtain the warning information, including: warning information, violation correction information, and positioning information;

[0181] The beneficial effects of the above technical solution are: by conducting comprehensive detection of human portraits, the accuracy of detecting violations when boarding and alighting a passenger vehicle is improved; by detecting human portrait violations and violation levels, the safety of the passenger vehicle violation response plan is improved.

[0182] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for analyzing illegal boarding and disembarking of passengers on passenger vehicles, characterized in that: The following steps are involved: Step S01: Screening the passenger transport monitoring data for vehicle violations to obtain the illegal boarding and disembarking feature values; Step S02: Perform feature analysis based on the illegal boarding and disembarking feature values ​​to construct a vehicle violation model; Step S03: generating output value similarity by performing model comparison calculation on the vehicle violation model; Step S04: performing deviation identification according to the output value similarity to obtain the information of the vehicle to be inspected; Step S05: Performing violation detection on the vehicle to be inspected according to the information of the vehicle to be inspected, and obtaining the vehicle violation information; Wherein, the model comparison calculation includes the following steps: Step 1: Obtain normalized data by normalizing the vehicle violation model data; wherein, The normalization process maps the conversion result to a preset interval by normalizing the deviation of the vehicle violation model data; wherein, The deviation standardization includes: linearly transforming the vehicle violation model data through a preset conversion function; Step 2: Transmit the normalized data to the twin network for mapping analysis and processing to generate twin feature space data; wherein, The twin network includes: extracting dual input data of normalized data, inputting it into the neural network, and performing mapping processing; wherein, The dual input data includes: previous frame data and next frame data; Step 3: perform output calculation on the twin feature space data, obtain the output value similarity, and make a judgment; wherein, When the output value similarity is less than or equal to a preset threshold, it is a normal vehicle behavior; When the output value similarity is greater than a preset threshold, it is abnormal vehicle behavior.

2. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The step S01 comprises: By classifying the passenger monitoring data according to the data function, the monitoring data category is obtained; among them, The passenger monitoring data includes: checkpoint data and video structured data; The monitoring data categories include: digital category, speed category, time category, and behavior category; By performing category feature extraction on each detection data category, a category feature value of each category is generated, and passenger boarding and disembarking analysis is performed on the category feature value to obtain a feature value of illegal passenger boarding and disembarking.

3. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The step S02 comprises: By modeling and screening the characteristic values ​​of illegal boarding and disembarking, the characteristic information of the illegal model is obtained; The violation model feature information includes: location information, node information, edge information, and interval information; According to the violation model feature information, a subject model is constructed to generate a vehicle violation model.

4. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The step S04 comprises: By matching the corresponding calculation method according to the output value similarity, performing deviation calculation, obtaining the deviation value, and making a judgment; wherein, The deviation calculation includes: comparing the output value with a preset output comparison data set according to the corresponding calculation method to obtain a comparison deviation value; When the deviation value is greater than a preset threshold, it is considered as an excessive deviation, and the vehicle is determined to be a to-be-detected vehicle, and the information of the to-be-detected vehicle is obtained; When the deviation value is less than or equal to a preset threshold, it is a normal deviation.

5. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The step S05 comprises: The first screening information is obtained by screening the key data of the vehicle information to be inspected; wherein, The key data screening includes: mobile status screening and path screening; Extract vehicle information from the first screening information to obtain a first vehicle to be inspected, and perform violation detection on the first vehicle to be inspected to obtain vehicle violation information; wherein, The violation detection includes: passenger vehicle detection and passenger vehicle license plate recognition.

6. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 5, characterized in that: The passenger vehicle detection comprises the following steps: Step S10: Obtain information feature layers by performing feature stratification on the vehicle information to be detected; wherein, The feature stratification includes: scale stratification, cost stratification, and training stratification; Step S20: extracting feature information of each information feature layer respectively according to the preset multi-scale feature learning, and generating information feature nodes corresponding to the information feature layer; Step S30: performing subspace mapping on the information feature layer and the corresponding information feature node, obtaining a mapping result, and judging the mapping result; wherein, When the mapping result is within a preset threshold range, it is a compliant passenger vehicle; When the mapping result is not within the preset threshold range, it is an illegal passenger vehicle, the illegal passenger vehicle information is obtained, and the preset violation response process is performed.

7. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 5, characterized in that: The passenger vehicle license plate recognition comprises the following steps: Step S100: Perform convolution feature extraction through a preset detection convolutional neural network to obtain a feature map; wherein, The convolution feature extraction includes: convolution, activation function, and building a pooling layer; Step S200: Performing a preset feature network analysis process according to the feature graph to generate a passenger vehicle detection frame; wherein, The feature network analysis process includes: region proposal network processing, fully connected layer processing; wherein, The region proposal network processing includes: generating a candidate region, obtaining a candidate region node, determining the node attributes of the candidate region, performing bounding box regression processing according to the node attributes, and determining a first candidate region; wherein, The node attributes include: foreground node and background node; The fully connected layer processing includes: extracting comprehensive information based on the first area to be selected to generate a feature map of the area to be selected, performing frame regression calculation based on the feature map of the area to be selected to obtain a passenger vehicle detection frame; Step S300: Performing a passenger vehicle violation detection according to the passenger vehicle detection border to obtain passenger vehicle violation information.

8. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The violation detection includes: By performing segmented detection on passenger vehicles, segmented detection information is obtained; wherein, The segment detection includes: time segment detection and route segment detection; According to the segment detection information, segment similarity analysis is performed to obtain segment similarity values ​​and make judgments; wherein, The segmented similarity analysis is performed by calculating and analyzing the similarity between each segment of information and the previous segment of information; When the segment similarity value is within a preset threshold range, it is a normal segmentation; When the segment similarity value is not within the preset threshold range, it is an abnormal segment and an abnormal analysis process is performed.

9. The analysis method based on illegal boarding and disembarking of passengers on passenger vehicles as claimed in claim 1, characterized in that: The violation detection also includes: By detecting the portrait violation behavior of the vehicle violation information, the portrait violation data is obtained; wherein, The portrait violation behavior detection includes: boarding behavior detection and getting off the vehicle behavior detection; According to the portrait violation data and the preset violation database, data comparison processing is performed to obtain the violation level; wherein, The violation levels include: minor violation, ordinary violation, and serious violation; According to the violation level, a corresponding security warning is issued to obtain warning information; wherein, The safety warning includes: providing an online warning to the vehicle where the offending person is located according to the violation level, and obtaining warning information; The violating vehicle performs violation correction processing according to the warning information and generates violation correction information; The early warning information includes: warning information, violation correction information, and positioning information.

Images