A rail transit non-inductive ticket checking method and system based on spatiotemporal trajectory and biological feature fusion

Abstract

The application relates to the technical field of ticket selling equipment, and discloses a rail transit non-inductive ticket checking method and system based on space-time trajectory and biological feature fusion, which comprises the following steps: continuously acquiring the space coordinate sequence of a moving target in a ticket checking area and corresponding biological recognition feature groups; determining the distribution weight according to the running deflection angle of the moving target relative to a sensor, calculating a verification confidence index; establishing a ticket checking decision line, a first buffer zone and a second buffer zone with different physical lengths on both sides of the ticket checking decision line on a passing path; monitoring the displacement vector of the moving target relative to a decision domain, and driving a ticket checking transaction state machine to migrate; and the application eliminates the mapping ambiguity between identity evidence dispersion and physical displacement continuity through an asymmetric logic buffer zone and a state machine migration mechanism.

Description

Technical Field

[0001] This invention belongs to the field of ticketing equipment technology, and in particular relates to a contactless ticket checking method and system for rail transit based on the fusion of spatiotemporal trajectory and biometric features. Background Technology

[0002] Currently, in automated fare collection systems for rail transit, open-style contactless ticketing improves station throughput by eliminating physical turnstiles. Contactless ticketing typically uses depth sensing units and visual acquisition components to capture the movement trajectory and biometrics of passengers, establishing a binding relationship between the identified passenger identity information and specific passage events. This mainstream technology provides a foundational path for building contactless urban rail transit ticketing systems. However, simply relying on increasing the performance of sensing hardware or iterating general recognition algorithms cannot solve the logical disconnect between identity attribution and physical trajectory. For example, the Chinese invention patent with authorization announcement number CN119785387B... This paper discloses a ground-air cooperative target recognition method based on self-evolving visual cue learning. It improves recognition accuracy in cross-platform environments by decoupling perspective feature extraction and feature refinement. However, in the rail transit billing scenario, simply improving recognition stability cannot eliminate the mapping ambiguity caused by the discreteness of identity evidence. It lacks in-depth mining of the physical displacement delay characteristics of the passing object. Faced with conditions such as random swinging near the decision perimeter, centroid shift caused by luggage occupying space, and trajectory adhesion under high-density passenger flow, the recognition result and the triggering of billing action are in a weak correlation state, resulting in the inversion of billing subject and object, leading to frequent false transaction records and failing to meet the legal certainty and evidence integrity requirements of billing settlement.

[0003] However, as a typical high-density passenger flow application scenario, rail transit stations have extremely high requirements for the legal certainty of billing transactions. In actual engineering applications, physical displacement is a continuous evolutionary process. However, due to factors such as fluctuations in ambient light and shadow, random object posture, or physical occlusion between people, the collection of biometric features presents a discrete, discontinuous, and random distribution. In order to maintain billing accuracy, existing technologies tend to over-rely on instantaneous recognition accuracy. This design mode, while using high computing resources to combat signal noise when dealing with high-density passenger flow, still cannot avoid billing disputes caused by the breakage of the evidence chain. To address the above challenges, simply increasing the sampling frequency or introducing complex recognition algorithms, or other linear improvement paths, will cause a significant increase in data processing load and make it difficult to eliminate the ambiguity between discrete identity evidence and continuous physical trajectory. This triggering mode that relies on instantaneous judgment results makes it easy for the system to generate a logical reversal of billing subject and object when facing stressful situations such as passengers lingering for a short time, physical adhesion of objects, and trajectory intersection collisions, thereby undermining the business foundation of the ticketing system as the arbitrator of billing logic.

[0004] Therefore, how to achieve asynchronous association and logical latching of discrete identity features in continuous spatiotemporal coordinate vectors, and to construct a deterministic arbitration mechanism for complex passage intentions, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features, comprising the following steps:

[0006] Step S1: Real-time acquisition of continuous spatial coordinate sequence of moving targets within the ticket inspection area, and extraction of biometric feature groups aligned with the continuous spatial coordinate sequence in the sampling time sequence;

[0007] Step S2: Calculate the similarity probability value between each sub-feature in the biometric feature group and the preset account feature database, and calculate the verification confidence index that represents the certainty of identity attribution based on the assigned weight determined by the travel angle of the moving target relative to the sensor.

[0008] Step S3: Establish a ticket inspection decision line and a first buffer zone and a second buffer zone distributed on both sides of the ticket inspection decision line on the passage path. Set the first buffer zone to be located on the entrance side of the ticket inspection decision line and its physical length along the direction of travel to be greater than the physical length of the second buffer zone located on the exit side.

[0009] Step S4: Monitor the relative displacement vector of the moving target with respect to the first buffer, the second buffer, and the ticket check decision line, and drive the ticket check transaction state machine to perform state transition: When the moving target passes through the first buffer and its centroid position crosses the ticket check decision line, and the verification confidence index is not lower than the preset deduction confidence threshold, the ticket check transaction state machine jumps to the billing compliance state and generates a deduction instruction message; when the moving target's centroid position crosses the ticket check decision line and the verification confidence index is lower than the deduction confidence threshold, the ticket check transaction state machine jumps to the transaction suspension state, locks the continuous spatial coordinate sequence until the complete biometric data is obtained in the subsequent tracking period to update the verification confidence index.

[0010] Preferably, in step S2, the calculation rules for the verification confidence index are as follows: ,in, To verify the confidence level indicators; This represents the total number of independent feature dimensions contained in the biometric feature set, and It is a positive integer; For the weight allocation, the corresponding number of... Weight coefficients of dimensional features; For the similarity probability value corresponding to the first The probability components of a dimensional feature.

[0011] Preferably, in step S3, the displacement delay interval formed by the physical length of the first buffer is used to filter the random swaying displacement generated by the moving target at the edge of the ticket check decision line, so as to prevent state jump oscillation.

[0012] Preferably, step S1 further includes the following sub-steps: monitoring the envelope volume value of the moving target point cloud corresponding to the continuous spatial coordinate sequence; when the envelope volume value of the moving target point cloud exceeds the preset single object volume threshold, transmitting a modulated infrared detection sequence to the ticket checking area; collecting phase offset data formed by reflection from the surface of the moving target, and performing logical decoupling of multiple overlapping moving targets from the continuous spatial coordinate sequence based on the spatial gradient distribution of the phase offset data.

[0013] Preferably, in step S4, the ticket checking transaction state machine performs the following operations while the transaction is suspended: Step S501, maintain closed-loop tracking of the moving target and open the auxiliary acquisition window to acquire the lateral or back area features of the moving target; Step S502, use the acquired lateral or back area features to incrementally correct the biometric feature group and recalculate the verification confidence index; Step S503, when the corrected verification confidence index reaches the deduction confidence threshold, trace back the moment when the moving target crosses the ticket checking decision line and resend the deduction instruction message.

[0014] Preferably, the acquisition of the continuous spatial coordinate sequence includes: acquiring a depth image stream of the ticket checking area using a depth sensing camera, with the sampling frequency set to not less than 30Hz; extracting the centroid coordinates of the moving target in the depth image stream, connecting the centroid coordinates of the moving target in each sampling frame to generate a continuous spatial coordinate sequence; calculating the displacement vector between adjacent sampling frames in the continuous spatial coordinate sequence, and determining the physical intrusion depth of the moving target relative to the first buffer or the second buffer.

[0015] Preferably, the method further includes the following steps: Step S701, calculating the path curvature continuity parameter of the continuous spatial coordinate sequence; Step S702, when the trajectory of the moving target is interrupted and reappears in the sensor coverage blind zone, comparing the path curvature continuity parameter before and after the interruption with the evolution trend of the verification confidence index, if the deviation is within the preset range, then performing the associated transparent transmission of the ticket check mark to realize trajectory logic self-repair.

[0016] Preferably, it further includes: identifying accompanying attributes in a continuous spatial coordinate sequence, including the geometric deviation value of the carry-on relative to the centroid of the moving target; and automatically calibrating the effective centroid position in the continuous spatial coordinate sequence based on the geometric deviation value to compensate for the trajectory displacement deviation caused by the luggage occupying space.

[0017] Preferably, the biometric feature group includes facial features, gait structure features, and clothing texture features. The weights are allocated based on real-time linear compensation adjustment of ambient light intensity data fed back by the photosensitive element. After step S4, the following steps are also included: Step S1001, recording the transition path of the ticket checking transaction state machine, the verification confidence index sequence, and the generation time of the deduction instruction message; Step S1002, encapsulating the recorded transition path, verification confidence index sequence, and generation time into a ticket checking and evidence storage data block and storing it in local memory as the basis for ticket auditing.

[0018] A contactless ticket checking system for rail transit based on the fusion of spatiotemporal trajectory and biometric features is characterized by comprising a data acquisition module, a feature processing module, a spatial decision-making module, and a ticket checking transaction module.

[0019] The data acquisition module is used to acquire the continuous spatial coordinate sequence of moving targets within the ticket inspection area in real time, and simultaneously extract the biometric feature group that is aligned with the continuous spatial coordinate sequence in the sampling time sequence.

[0020] The feature processing module is used to calculate the verification confidence index based on the similarity probability values ​​of each feature in the biometric feature group and the preset account feature database, and combined with the assigned weights determined by the travel angle of the moving target relative to the sensor. The feature processing module is also used to calculate the path curvature continuity parameter of the continuous spatial coordinate sequence, and to identify the geometric deviation value of the carried object relative to the centroid of the moving target in order to calibrate the effective centroid position.

[0021] The spatial decision module is used to establish a decision domain with physical displacement delay characteristics on the passage path, including a ticket inspection decision line and a first buffer zone and a second buffer zone distributed on both sides of the ticket inspection decision line. The first buffer zone is set to be located on the entrance side of the ticket inspection decision line and its physical length along the direction of travel is greater than the physical length of the second buffer zone located on the exit side.

[0022] The ticket checking module monitors the relative displacement vector of the moving target with respect to the decision domain, driving the ticket checking state machine to perform state transitions; it verifies the confidence index when the moving target passes through the first buffer zone and its centroid crosses the ticket checking decision line. If the amount is not lower than the preset deduction confidence threshold, the system will switch to the billing compliance status and generate a deduction instruction message; the confidence index will then be verified. When the data falls below the deduction confidence threshold, the system jumps to a transaction suspension state and locks the continuous spatial coordinate sequence until the complete biometric data is obtained.

[0023] Compared with existing technologies, the contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features has the following advantages:

[0024] 1. In the contactless ticket checking of rail transit, a ticket status determination mechanism based on asynchronous probability weight accumulation is established. The discrete biometric features captured in the passage area are dynamically allocated to the corresponding spatiotemporal coordinate vectors according to their spatial correlation through the weight mapping matrix. This realizes the logical decoupling of the biometric recognition process and the billing trigger action in the time dimension. By utilizing the displacement continuity of the trajectory in the sensing area, spatiotemporal evolution evidence is provided for the identity certainty in the verification interval. This eliminates the failure of the deduction account mapping caused by instantaneous occlusion or drastic fluctuations in light and shadow, and ensures the integrity and non-repudiation of billing transactions of ticketing equipment in high-density passenger flow scenarios.

[0025] 2. A billing arbitration logic centered on the convergence of ticketing transaction confidence is established. By continuously monitoring the dynamic evolution of transaction entropy within the verification interval, the billing identifier is locked in a legally confident state. The system only triggers a deduction instruction when the trajectory crosses the decision line and the confidence level of its associated identity features meets the preset legal threshold for billing. If the confidence level does not meet the standard, the transaction is automatically suspended and guided to an asynchronous review process. This non-linear deterministic convergence judgment eliminates the risk of erroneous deductions caused by fluctuations in identity mapping, and builds a robust billing business firewall for open gateless access.

[0026] 3. Construct an identity chain maintenance mechanism based on physical motion characteristic frequency and path inertia. In the case of missing visual features in the core verification area, the mechanism uses the periodic fluctuation frequency of the object's centroid in the vertical axis direction to assist in the logical latching of the billing identifier. This mechanism reuses the displacement data originally used for trajectory generation as motion rhythm features. By comparing the consistency between the real-time fluctuation frequency and the historical records, the system can achieve deterministic transmission of the billing identifier in the visual blind zone without the need for additional sensors. Combined with path curvature continuity analysis, the system has the logical self-repair capability when the object's topological distance is reduced or the trajectory is interfered with or collided, ensuring the determinism of the billing subject's judgment result. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the contactless ticket checking transaction processing flow and state transition of the present invention;

[0028] Figure 2 This is a curve showing the relationship between the multidimensional biometric feature fusion weights and the travel angle of this invention;

[0029] Figure 3 This is a fishbone diagram of the logical architecture of the contactless ticket checking system based on the spatiotemporal trajectory and feature fusion of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0031] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, low, lateral, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated.

[0032] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal communication between two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0033] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0034] This invention provides a contactless ticket checking method and system for rail transit based on the fusion of spatiotemporal trajectory and biometric features. It comprises a data acquisition module, a feature processing module, a spatial decision-making module, and a ticket checking transaction module. These modules communicate via a data bus to jointly complete passenger identity verification and ticket payment settlement. In the ticket checking area, high-density passenger flow often results in discrete and discontinuous biometric feature collection, making it difficult to establish a stable mapping with continuous travel trajectories. To establish legal certainty in billing transactions, the data acquisition module utilizes a depth-sensing camera to acquire depth image streams of the ticket checking area, with a sampling frequency of no less than [missing information]. The system extracts the centroid coordinates of moving targets from the depth image stream, connects the centroid coordinates of moving targets in each sampled frame to generate a continuous spatial coordinate sequence, and the feature processing module extracts biometric feature groups aligned with the continuous spatial coordinate sequence in the sampling time sequence. The biometric feature groups include facial features, gait structure features, and clothing texture features. The feature processing module determines the weights to be assigned based on the azimuth angle of the moving target relative to the sensor. Calculate the verification confidence index Verification confidence index The calculation rules are as follows: ,in, To verify the confidence level indicators. This represents the total number of independent feature dimensions contained in the biometric feature set, and It is a positive integer. For the weight allocation, the corresponding number of... Weight coefficients of dimensional features For the similarity probability value corresponding to the first The probability components of a dimensional feature.

[0035] Select travel angle exist to Scope For discrete sampling points of the step length, obtain sample sequences of facial features, gait structure features, and clothing texture features corresponding to each sampling point, and calculate the information entropy of each dimension of features at the corresponding skew angle. Information entropy Representing the uncertainty of feature matching, based on information entropy The weight coefficient is determined by the proportion of the reciprocal in the sum of the reciprocals of the feature dimensions. Weighting coefficient The calculation formula is as follows: ,in This represents the total dimension of biometric features. For the first The information entropy of the feature at the current deflection angle is used to establish information about the deflection angle. The weight mapping matrix is ​​used to obtain the slant angle of the moving target from the feature processing module. Then, the weight mapping matrix is ​​retrieved and linear interpolation is performed to output the weight coefficients for instantaneous pose matching. And calculate the verification confidence index. Random swaying of objects often occurs at the edge of the ticket inspection decision line, which can easily cause oscillations in the billing status. The spatial decision module establishes a decision domain along the travel path that includes the ticket inspection decision line, a first buffer zone, and a second buffer zone. The first buffer zone is set to be located on the entrance side of the ticket inspection decision line, and its physical length along the travel direction is defined as follows: Greater than the physical length of the second buffer zone located on the outbound side The spatial decision module utilizes the physical length of the first buffer. The resulting displacement delay interval filters out displacement fluctuations caused by the moving target at the edge of the ticket inspection decision line, preventing state transition oscillations. The ticket inspection transaction module monitors the relative displacement vector of the moving target relative to the decision domain and drives the ticket inspection transaction state machine to perform state transitions. When the moving target crosses the first buffer zone, its centroid position crosses the ticket inspection decision line, and the confidence index is verified. Not lower than the preset deduction confidence threshold At that time, the ticket checking transaction state machine jumps to the billing compliance state and generates a deduction instruction message, eliminating the ambiguity between the discreteness of identity evidence and the continuity of physical displacement.

[0036] For verification confidence index Below the charge confidence threshold In nondeterministic scenarios, the ticket checking module executes the data completion program in the auxiliary data collection window to determine the physical displacement of the moving target after its centroid crosses the ticket checking decision line. An auxiliary acquisition window is defined within the range. The feature processing module acquires the texture feature vector of the back area of ​​the moving target within this window, and then performs the acquisition within a preset range. Within the tracking step, incremental weight adjustments are performed on the biometric feature groups until the adjusted verification confidence index is reached. Greater than or equal to the deduction confidence threshold The billing transaction state machine is triggered to transition from the transaction suspension state to the billing compliance state, thereby ensuring the completion of the evidence loop for billing transactions under the condition of restricted biometric features. In the case of multiple people overlapping or carrying luggage, the system performs point cloud envelope processing to establish the uniqueness of the billing subject. The data acquisition module monitors the volume value of the point cloud envelope of the moving target corresponding to the continuous spatial coordinate sequence. When the volume value exceeds the preset single object volume threshold, the modulated infrared detection sequence is emitted to the ticket inspection area. The data acquisition module collects the phase offset data formed by the reflection of the target surface and performs logical decoupling of the overlapping moving targets from the continuous spatial coordinate sequence according to the spatial gradient distribution of the phase offset data. The feature processing module identifies the accompanying attributes in the coordinate sequence, namely the geometric deviation value of the carry-on relative to the centroid of the moving target, and automatically calibrates the effective centroid position according to the geometric deviation value to compensate for the trajectory displacement deviation caused by the luggage occupying space.

[0037] Get Point cloud envelope volume data of adult samples of different heights and body types under standard walking postures were used to calculate confidence intervals for the volume data using a normal distribution model, and confidence levels were selected. The upper limit of the volume at a location determines the threshold for the volume of a single object. When real-time monitoring of the envelope volume value of moving target point cloud satisfy At that time, it was determined that multiple targets were physically stuck together in the ticket checking area, and the data acquisition module was driven to transmit the modulation frequency. The infrared detection sequence collects phase shift data formed by reflections from the target surface. Based on the spatial gradient distribution of the phase shift data in the polar coordinate system, the depth fault boundary is identified, and multiple moving targets that are adhered together are logically decoupled in a continuous spatial coordinate sequence. To establish the logical continuity of billing transactions before and after the sensor coverage blind zone, a trajectory curvature continuity calculation procedure is executed. The feature processing module extracts the end coordinate subsequence of the moving target before entering the blind zone and the starting coordinate subsequence after leaving the blind zone, and calculates the curvature value between adjacent spatial sampling points. curvature value The calculation formula is as follows: ,in, The curvature value, , These are the first derivatives of the spatial coordinate axis components of the moving target with respect to the sampling number. , These are the second derivatives of the spatial coordinate axis components with respect to the sampling sequence number. The feature processing module compares the curvature change trends of the two subsequences at the temporal junction. When the deviation of the curvature value is less than the preset curvature fluctuation threshold... And the verification confidence index corresponding to the two trajectories The mean shift is less than At that time, the system performs the associated transparent transmission of the ticket check identifier, realizing the stability of the billing transaction logic in the state of perceived interruption.

[0038] Example 1: In the peak operation scenario of a rail transit hub station, when the system faces an instantaneous passenger flow of up to [number] passengers per minute... The number of people and the physical distance between adjacent objects is less than Under environmental pressure, the physical overlap between human bodies and the random pauses and swaying of passengers around the ticket inspection decision line cause the biometric collection signals to exhibit high dispersion. This leads to ambiguity in the binding of identity evidence and movement trajectory at key points in the traditional ticket inspection logic at the fare calculation stage. To address the recognition obstruction caused by the aforementioned passenger flow intermingling, the spatial decision module establishes heterogeneous first and second buffer zones along the passage path. The physical length of the first buffer zone located on the entrance side is... Set as The physical length of the second buffer zone located on the outbound side Set as By utilizing the physical length of the first buffer The system creates a physical displacement delay interval, which restricts the displacement fluctuations of the object at the edge of the ticket check decision line to the first buffer, thereby achieving physical filtering of invalid billing intentions.

[0039] Within this spatially constrained path, the feature processing module synchronously executes a dynamic weight allocation procedure. When facial features are visually obscured due to dense traffic, the system allocates weights based on the object's travel angle relative to the depth sensing camera. The feature processing module shifts the collected biometric feature groups towards gait structure and clothing texture features, performs anonymization mapping on the resulting feature groups, generates a feature identifier code associated with the user account using a one-way hash function, and then calculates the verification confidence index based on the continuous spatial coordinate sequence generated in the previous sampling period. Verification confidence index The calculation formula is as follows: ,in, To verify the confidence level indicators. This represents the total number of independent feature dimensions contained in the biometric feature set, and It is a positive integer. For the weight allocation, the corresponding number of... Weight coefficients of dimensional features For the similarity probability value corresponding to the first The probability components of a dimensional feature.

[0040] When the centroid vector of the moving target completely passes through the first buffer zone and crosses the ticket check decision line, and the verification confidence index is met... Not lower than the deduction confidence threshold When the ticket checking transaction module drives the ticket checking transaction state machine to jump from the transaction suspension state to the billing compliance state, it generates a deduction instruction message. For objects carrying luggage, the system calculates the geometric deviation value of the luggage relative to the centroid of the moving target and outputs the calibrated valid centroid coordinates as the trigger basis to eliminate the trajectory center offset caused by the luggage occupying space and establish the uniqueness of the billing sovereignty. Through the application of the correlation between the asymmetric spatial buffer and the feature weight distribution logic, the system establishes a deterministic mapping between the discrete identity evidence sequence and the continuous physical trajectory displacement. Without increasing the physical gate hardware, it solves the technical conflict between passage efficiency and billing rigor in the open passage scenario. The system encapsulates the transition path of the ticket checking transaction state machine, the verification confidence index sequence and the deduction instruction message into a ticket checking evidence storage data block and stores it in the local memory to achieve evidence self-consistency of the billing transaction in the ticketing system.

[0041] Example 2: When the physical length is On the open-access corridor verification platform, using a measurement resolution of And the sampling frequency is A depth-sensing camera captures point cloud data of moving targets and simulates passenger flow distribution characteristics during peak hours at a rail transit hub. The signal-to-noise ratio of the actively superimposed signal source in the experiment is [value missing]. Gaussian white noise was used to verify the accuracy of the billing transaction processing under visual occlusion and background fluctuation interference; the data acquisition module acquired the depth image stream and extracted the centroid coordinates of moving targets, and connected the centroid coordinates of moving targets in each sampling frame to form a continuous spatial coordinate sequence; regarding the physical length of the first buffer... The technical balance in this design lies in the filtering effectiveness of displacement fluctuations at the edge of the ticket checking decision line and the immediacy of the billing response; if the physical length Values ​​lower than A slight backward movement or pause of the object's centroid at the moment of crossing the line will trigger a state transition in the billing state machine between the billing compliance state and the transaction suspended state; if the physical length Values ​​exceeding This results in a physical delay in generating the fare deduction instruction exceeding the real-time specifications for rail transit ticketing auditing; in this test condition, for walking speed distribution in to The physical length of the first buffer zone is selected based on the passenger flow groups between them. for The physical length of the second buffer for And will verify the confidence index Deduction confidence threshold Set as The feature processing module extracts facial features, gait structure features, and clothing texture features in real time. It then performs anonymization mapping on the collected biometric feature groups and uses a one-way hash function to generate feature identifiers associated with user accounts. To verify the synergistic effect and the rationality of parameter boundaries, experiments were conducted using the present invention's sample group, a feature-deficient control group, a symmetric buffer control group, and a group containing physical length... Values or The experiment included multiple groups, such as the control group exceeding the permitted range; the table below shows the percentage of the human torso area obscured in different experimental groups. It also exhibits performance under conditions of random perimeter oscillation.

[0042] Table 1: Comparison of ticket checking performance of different test groups under obstructed conditions

[0043]

[0044] Referring to Table 1, in complex occlusion environments, the sample group of this invention dynamically adjusts the weight allocation based on the object's travel angle through the feature processing module. To make the verification confidence index Maintain at The accuracy rate of billing sovereignty attribution reached a certain level. Verification confidence index The calculation formula is as follows: ,in, To verify the confidence level indicators. This represents the total number of independent feature dimensions contained in the biometric feature set, and It is a positive integer. For the corresponding number Weight coefficients of dimensional features For the corresponding number The similarity probability components of the features; comparative data shows that the feature-missing control group, due to the use of fixed weight allocation, cannot compensate for the loss caused by missing facial features, thus affecting the verification confidence index. Descending to The symmetric buffer control group lacks physical length. for The displacement delay interval cannot effectively filter out the random displacement of the object at the edge of the ticket inspection decision line, resulting in... Secondary pseudo-transaction record.

[0045] Further observation of the gradient performance changes at the boundaries of key parameters, when the physical length of the first buffer... Reduced to At that time, because the displacement delay interval was insufficient to cover the amplitude of the physical micro-vibrations generated by the passenger at the perimeter, the number of pseudo-transaction record triggers increased to [number missing]. Next; and when the physical length Increase to At that time, the deviation in the time of billing instruction generation increased to This indicates that the system response efficiency decreases due to excessively long physical displacement travel; the system monitors the volume of the moving target point cloud envelope corresponding to a continuous spatial coordinate sequence, and when a physical overlap occurs where two people are walking side by side, causing the volume value to exceed [a certain threshold], [the system's response efficiency decreases]. When the single-person object volume threshold is reached, a modulation frequency of is transmitted to the ticket checking area. The infrared detection sequence is analyzed; by analyzing the spatial gradient distribution of the surface reflection phase shift data, multiple overlapping moving targets are logically decoupled, and the geometric deviation value of the carrier relative to the centroid of the moving target in the coordinate sequence is identified. The calibrated effective centroid coordinates are output as the triggering basis.

[0046] Example 3: This example combines Figures 1 to 3 This paper describes a contactless ticket checking method and system for rail transit based on the fusion of spatiotemporal trajectory and biometric features. Figure 1As shown, step S1 involves acquiring the continuous spatial coordinate sequence of moving targets within the ticket checking area in real time, and extracting biometric feature groups aligned with the continuous spatial coordinate sequence in the sampling time sequence. Step S2 involves calculating the similarity probability value between each feature in the biometric feature group and the preset account feature database, and calculating the verification confidence index representing the certainty of identity attribution based on the weight determined by the travel angle of the moving target relative to the sensor. Step S3 involves establishing a ticket checking decision line and a first buffer and a second buffer distributed on both sides of the decision line on the passage path. The first buffer is set to be located on the entrance side and its physical length along the travel direction is greater than the physical length of the second buffer located on the exit side. Finally, in step S4, the relative displacement vector of the moving target relative to the buffer and the decision line is monitored to drive the ticket checking transaction state machine to migrate. The specific logic is as follows: when the centroid crosses the decision line and the verification confidence index meets the standard, the process jumps to the billing compliance state; when the index does not meet the standard, the process jumps to the transaction suspension state and locks the trajectory sequence.

[0047] like Figure 2 As shown, the horizontal axis represents the travel angle, ranging from 0° to 180°, and the vertical axis represents the probability or confidence level. The figure shows four key curves with different line types: the facial feature similarity probability curve, represented by the dashed line, maintains high values ​​at the direct angles of 0° and 180°, but drops to its lowest point at the lateral angle near 100°. Complementing this are the clothing texture feature similarity probability curve, represented by the dotted line, and the gait structure feature similarity probability curve, represented by the midpoint dashed line. Both of these curves exhibit relatively high values ​​in the middle lateral angle range. Through dynamic weight fusion, the verification confidence index C, represented by the solid line, maintains a stable and high level throughout the entire angle range, verifying the system's ability to maintain identity verification stability under different travel angles.

[0048] like Figure 3As shown, in the data acquisition dimension, a depth camera with a sampling frequency of no less than 30Hz is deployed in the system, and combined with a modulated infrared detection sequence to obtain high-precision physical perception data. In the spatial decision domain dimension, the system establishes an asymmetric buffer along the passage path, that is, the physical length L1 of the first buffer on the entrance side is set to be greater than the physical length L2 of the second buffer on the exit side. The physical displacement delay characteristics are used to filter out random swing interference at the edge of the ticket inspection decision line. In the biometric feature fusion dimension, the system extracts multimodal features including face, gait, and clothing texture, and uses a base The system employs a dynamic weighting mechanism for the travel angle to accurately calculate the verification confidence index C. In the complex scenario compensation dimension, the system implements trajectory curvature repair for blind spots, decoupling of overlapping targets using phase gradients, and baggage geometric deviation compensation to ensure the rigor of judgment in dense passenger flow or when carrying obstacles. Finally, in the ticket checking transaction state machine dimension, the system monitors the displacement vector of the moving target relative to the decision domain. When the centroid crosses the decision line and the verification confidence level meets the standard, the system triggers the deduction. If the confidence level is lower than the threshold, the system enters a transaction suspension and locking state, thus constructing a complete closed-loop ticket checking logic.

[0049] Example 4: When the system faces the extreme condition of three people walking side by side with no physical gaps between adjacent objects, the point cloud envelope volume value captured by the data acquisition module continuously exceeds [a certain value]. Because the topological structure of the moving target at this time exhibits a single connected component characteristic, the morphological clustering algorithm cannot segment the billing entity, resulting in the risk of misjudging that a single account carries multiple people's ticketing transactions. Therefore, when the system detects that the volume value exceeds the limit, it initiates an active detection procedure, ensuring that the transmission power density is not lower than... And the modulation frequency is The modulated infrared detection sequence acquires phase shift data formed by reflection from the target surface; the feature processing module executes a spatial segmentation algorithm based on phase gradient, establishes a polar coordinate system with the optical axis of the depth sensing camera as a reference, performs phase demodulation on each spatial sampling point within the detection area and extracts the time-of-flight information of each point, and calculates the phase spatial gradient vector between adjacent sampling points. Phase space gradient vector The calculation formula is as follows: ,in, The phase space gradient vector; The phase value obtained from demodulation; , , These are the spatial coordinate axis components; the feature processing module retrieves them. The abrupt change point of the modulus determines the depth fault boundary between physical entities, decoupling the three moving targets, which are physically bonded, from the continuous spatial coordinate sequence into three independent logical identifiers.

[0050] The system determines the single-person object volume threshold through offline calibration experiments. Select The sample group includes adult samples with different heights and body types. Point cloud envelope volume data of each sample were measured under standard walking postures, and confidence intervals were calculated using a Gaussian distribution model. A confidence level of 100% was selected. The upper limit of volume is used as the threshold for the volume of a single object. The rated value; in this embodiment, the single-person object volume threshold. Calibrated as When the real-time monitored envelope volume value satisfy When the system detects physical adhesion, it activates the infrared detection sequence. To address the accompanying luggage obstruction, the feature processing module identifies geometric deviation values ​​in the coordinate sequence. Using a subtraction algorithm, it removes the moving point data corresponding to the human gait template, identifying the remaining point cloud blocks as carried items. The feature processing module then calculates the horizontal vector of the carried item's center relative to the centroid of the moving target. The system uses a mass distribution prediction model of the carried objects and the human body to determine the effective centroid position to the horizontal vector. move in the opposite direction The vector magnitude is increased by a factor of 1 to achieve real-time compensation of the centroid coordinates. Based on the decoupled trajectory and calibrated centroid coordinates, the ticket checking transaction module drives the ticket checking transaction state machine to migrate to the billing compliance state, thus establishing the robustness of the billing sovereignty of the rail transit ticketing system.

[0051] Example 5: In the deployment phase of newly established rail transit stations, the system executes a biometric weight calibration procedure, selecting simulated objects along the ticket checking area. to Within range Samples are taken at discrete travel angles, and the similarity probability values ​​of facial features, gait structure features, and clothing texture features at each sampled angle are calculated. It also extracts the perceptual resolution of features in each dimension under different deflection angles to establish contribution components, thereby constructing the travel deflection angle. With weighting coefficients The mapping table enables the feature processing module to obtain the real-time travel angle of the moving target. Then, query the mapping table to match the weight coefficients. And calculate the verification confidence index. The sum of the weighting coefficients satisfies the following formula: ,in, These are the weighting coefficients. This represents the total number of independent feature dimensions contained in a biometric feature group.

[0052] After connecting to the ticketing network, the system executes a de-identification process. The feature processing module uses a one-way hash function to convert the original image features and motion vector features into a binary encrypted identifier associated with the user account, and stores this encrypted identifier in the temporary register of the ticket checking module. The verification confidence index is then used when the centroid vector of the moving target crosses the first buffer and the line is crossed. Not lower than the deduction confidence threshold At that time, the ticket checking module sends an encrypted identifier and the corresponding deduction instruction message to the back-end settlement server. After the ticket checking state machine completes the reset procedure from the billing compliance state to the initial state, the system automatically clears the feature cache data in the current tracking period to establish the privacy and security boundaries of the ticket checking data.

[0053] Example 6: During the deployment phase of newly established rail transit stations, the system executes a weight allocation matrix... The pre-deployment calibration procedure is adapted to specific sensor installation heights and lighting conditions by selecting the travel angle. exist to Within the range A sampling sequence is established for the stride length to obtain facial features, gait structure features, and clothing texture features at each sampling angle. A set of biometric sampling data was collected, and the similarity probability values ​​of each dimension feature at the corresponding angle were calculated. With perceptual entropy Perceived entropy value To characterize the statistical uncertainty in the feature matching process, the system utilizes perceptual entropy values. The inverse proportion determines the weight distribution to achieve weight allocation for reliable feature dimensions. Weight coefficients of dimensional features The calculation formula is as follows: ;in, These are the weighting coefficients. For the first The perceptual entropy value of a dimensional feature at a specific travel angle. To determine the total number of independent feature dimensions within a biometric feature set, a mapping model was established between the travel angle in physical space and the weight coefficients. This model enables the feature processing module to automatically adjust the weight distribution to maintain the verification confidence index when lateral traffic causes an increase in the facial feature perception entropy. The validity of the judgment.

[0054] In response to the physical envelope centroid shift caused by luggage carried by passengers, the system applies a mass centroid compensation coefficient. The on-site verification procedure selects physical volume distribution in to Standard test suitcases in the interval were subjected to passability tests, and the horizontal vector of the suitcase's geometric center relative to the centroid of the moving target was extracted from each test sample. A linear compensation model was established based on the deviation of the measured trajectory from the human anatomical center. Under the operating conditions maintained in this deployment instance, the mass center of gravity compensation coefficient was... Set as Effective centroid offset The calculation formula is as follows: ;in, For effective centroid offset, This is the mass center of gravity compensation coefficient. Let the horizontal vector of the object's center relative to the moving target's center of mass be determined by the effective center of mass offset. The corrected valid centroid coordinates serve as the basis for billing triggering, eliminating trajectory displacement deviations caused by luggage occupying space and ensuring the consistency of the ticket checking transaction state machine's judgments within the passage path decision domain.

[0055] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the scope of protection of this application.

Claims

1. A contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features, characterized in that, Includes the following steps: Step S1: Real-time acquisition of continuous spatial coordinate sequence of moving targets within the ticket inspection area, and extraction of biometric feature groups aligned with the continuous spatial coordinate sequence in the sampling time sequence; Step S2: Calculate the similarity probability value between each feature in the biometric feature group and the preset account feature database, and calculate the verification confidence index representing the certainty of identity attribution based on the assigned weights determined by the travel angle of the moving target relative to the sensor; wherein, the travel angle is selected. exist to Scope For discrete sampling points of the step length, obtain sample sequences of facial features, gait structure features, and clothing texture features corresponding to each sampling point, and calculate the information entropy of each dimension of features at the corresponding skew angle. Information entropy Representing the uncertainty of feature matching, based on information entropy The weight coefficient is determined by the proportion of the reciprocal in the sum of the reciprocals of the feature dimensions. Weighting coefficient The calculation formula is as follows: ,in This represents the total dimension of biometric features. For the first The information entropy of the dimensional feature at the current angle; Step S3: Establish a ticket inspection decision line and a first buffer zone and a second buffer zone distributed on both sides of the ticket inspection decision line on the passage path. Set the first buffer zone to be located on the entrance side of the ticket inspection decision line and its physical length along the direction of travel to be greater than the physical length of the second buffer zone located on the exit side. Step S4: Monitor the relative displacement vector of the moving target with respect to the first buffer, the second buffer, and the ticket check decision line, and drive the ticket check transaction state machine to perform state transition: When the moving target passes through the first buffer and its centroid position crosses the ticket check decision line, and the verification confidence index is not lower than the preset deduction confidence threshold, the ticket check transaction state machine jumps to the billing compliance state and generates a deduction instruction message; when the moving target's centroid position crosses the ticket check decision line and the verification confidence index is lower than the deduction confidence threshold, the ticket check transaction state machine jumps to the transaction suspension state, locks the continuous spatial coordinate sequence until the biometric data is obtained in the subsequent tracking period to update the verification confidence index; In step S4, the ticket checking transaction state machine performs the following operations while the transaction is suspended: Step S501, maintain closed-loop tracking of the moving target and open the auxiliary acquisition window to acquire the lateral or back area features of the moving target; Step S502, use the acquired lateral or back area features to incrementally correct the biometric feature group and recalculate the verification confidence index; Step S503, when the corrected verification confidence index reaches the deduction confidence threshold, trace back the moment when the moving target crosses the ticket checking decision line and resend the deduction instruction message.

2. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features according to claim 1, characterized in that, In step S2, the calculation rules for the verification confidence index are as follows: ,in, To verify the confidence level indicators; This represents the total number of independent feature dimensions contained in the biometric feature set, and It is a positive integer; For the weight allocation, the corresponding number of... Weight coefficients of dimensional features; For the similarity probability value corresponding to the first The probability components of a dimensional feature.

3. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features as described in claim 1, characterized in that, In step S3, the displacement delay interval formed by the physical length of the first buffer is used to filter the random swaying displacement generated by the moving target at the edge of the ticket check decision line, so as to prevent state jump oscillation.

4. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features as described in claim 1, characterized in that, Step S1 also includes the following sub-steps: monitoring the envelope volume value of the moving target point cloud corresponding to the continuous spatial coordinate sequence; when the envelope volume value of the moving target point cloud exceeds the preset single object volume threshold, transmitting a modulated infrared detection sequence to the ticket checking area; collecting phase offset data formed by reflection from the surface of the moving target, and performing logical decoupling of multiple overlapping moving targets from the continuous spatial coordinate sequence based on the spatial gradient distribution of the phase offset data.

5. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features according to claim 1, characterized in that, The acquisition of the continuous spatial coordinate sequence includes: acquiring a depth image stream of the ticket checking area using a depth sensing camera, with the sampling frequency set to no less than 30Hz; extracting the centroid coordinates of the moving target in the depth image stream, connecting the centroid coordinates of the moving target in each sampling frame to generate a continuous spatial coordinate sequence; calculating the displacement vector between adjacent sampling frames in the continuous spatial coordinate sequence, and determining the physical intrusion depth of the moving target relative to the first buffer or the second buffer.

6. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features as described in claim 1, characterized in that, It also includes the following steps: Step S701: Calculate the path curvature continuity parameter of the continuous spatial coordinate sequence; Step S702: When the trajectory of the moving target is interrupted and reappears in the sensor coverage blind zone, compare the path curvature continuity parameter before and after the interruption with the evolution trend of the verification confidence index. If the deviation is within the preset range, then perform the associated transparent transmission of the ticket check mark.

7. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features according to claim 1, characterized in that, Also includes: Identify the accompanying properties in a continuous spatial coordinate sequence, including the geometric deviation of the object relative to the centroid of the moving target; The effective centroid position in the continuous spatial coordinate sequence is automatically calibrated based on the geometric deviation value to compensate for the trajectory displacement deviation caused by luggage occupying space.

8. The contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features according to claim 1, characterized in that, The biometric feature set includes facial features, gait structure features, and clothing texture features. The weights are allocated based on real-time linear compensation adjustment of ambient light intensity data fed back by the photosensitive element. After step S4, the following steps are also included: Step S1001, recording the transition path of the ticket checking transaction state machine, the verification confidence index sequence, and the generation time of the deduction instruction message; Step S1002, encapsulating the recorded transition path, verification confidence index sequence, and generation time into a ticket checking and evidence storage data block and storing it in local memory as the basis for ticket auditing.

9. A contactless ticket checking system for rail transit based on the fusion of spatiotemporal trajectory and biometric features, used to implement the contactless ticket checking method for rail transit based on the fusion of spatiotemporal trajectory and biometric features as described in claim 1, characterized in that, It includes a data acquisition module, a feature processing module, a spatial decision-making module, and a ticket checking module: The data acquisition module is used to acquire the continuous spatial coordinate sequence of moving targets within the ticket inspection area in real time, and simultaneously extract the biometric feature group that is aligned with the continuous spatial coordinate sequence in the sampling time sequence. The feature processing module is used to calculate the verification confidence index based on the similarity probability values ​​of each feature in the biometric feature group and the preset account feature database, and combined with the assigned weights determined by the travel angle of the moving target relative to the sensor. Among them, the travel deflection angle is selected. exist to Scope For discrete sampling points of the step length, obtain sample sequences of facial features, gait structure features, and clothing texture features corresponding to each sampling point, and calculate the information entropy of each dimension of features at the corresponding skew angle. Information entropy Representing the uncertainty of feature matching, based on information entropy The weight coefficient is determined by the proportion of the reciprocal in the sum of the reciprocals of the feature dimensions. Weighting coefficient The calculation formula is as follows: ,in This represents the total dimension of biometric features. For the first The information entropy of the feature at the current deflection angle; the feature processing module is also used to calculate the path curvature continuity parameter of the continuous spatial coordinate sequence, and to identify the geometric deviation value of the carried object relative to the centroid of the moving target to calibrate the effective centroid position; The spatial decision module is used to establish a decision domain with physical displacement delay characteristics on the passage path, including a ticket inspection decision line and a first buffer zone and a second buffer zone distributed on both sides of the ticket inspection decision line. The first buffer zone is set to be located on the entrance side of the ticket inspection decision line and its physical length along the direction of travel is greater than the physical length of the second buffer zone located on the exit side. The ticket checking module monitors the relative displacement vector of the moving target with respect to the decision domain, driving the ticket checking state machine to perform state transitions; it verifies the confidence index when the moving target passes through the first buffer zone and its centroid crosses the ticket checking decision line. If the amount is not lower than the preset deduction confidence threshold, the system will switch to the billing compliance status and generate a deduction instruction message; the confidence index will then be verified. When the value falls below the deduction confidence threshold, the system jumps to a transaction suspension state and locks the continuous spatial coordinate sequence until the complete biometric data is obtained. Furthermore, the ticket checking transaction state machine performs the following operations while in the transaction suspension state: maintains closed-loop tracking of the moving target and opens an auxiliary acquisition window to obtain the lateral or rear-facing features of the moving target; incrementally corrects the biometric feature set using the obtained lateral or rear-facing features and recalculates the verification confidence index; when the corrected verification confidence index reaches the deduction confidence threshold, it traces back to the moment the moving target crosses the ticket checking decision line and resends the deduction instruction message.

Images