Subway track automatic ticket checking method and automatic ticket checking device thereof

Abstract

The invention discloses a subway track automatic ticket checking method and an automatic ticket checking device thereof. The method specifically comprises the following steps: S1, feature acquisition; s2, performing feature fusion; s3, performing anomaly detection; the invention relates to the technical field of subway track automatic ticket checking. According to the subway track automatic ticket checking method and the automatic ticket checking device thereof, through cooperation of palm vein and gait recognition and a non-contact collection mode, the safety of a ticket checking system is improved, the false recognition rate is reduced, meanwhile, the accuracy of abnormal behavior detection is effectively improved, the self-learning ability is provided through a reinforcement learning algorithm, and the accuracy of ticket checking is improved. The ticket checking strategy can be automatically adjusted according to the real-time passenger flow and the operation state, optimal configuration multi-mode identification of resources is achieved, efficiency and safety are balanced, the crowding degree in the peak period can be reduced, and high-satisfaction travel experience is provided for passengers.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic ticket checking for subway tracks, and specifically to an automatic ticket checking method and an automatic ticket checking device for subway tracks. Background Art

[0002] As the backbone network of public transportation, the efficiency and security of the ticket checking system of the subway are directly related to the operation efficiency of the city. The existing ticket checking technologies mainly include:

[0003] Magnetic card / IC card ticket checking machines, which rely on physical media, are easy to lose and forge, and are prone to congestion during peak hours;

[0004] QR code / NFC ticket checking, although it realizes contactless passage, has problems of screen dependence and network latency;

[0005] Face recognition is restricted by masks and low-light environments, and there are also relatively large privacy disputes;

[0006] Fingerprint recognition requires contact operation, and there are health hazards and loopholes in live detection.

[0007] Moreover, the existing ticket checking machines have fixed ticket checking methods and cannot adapt to real-time passenger flow changes. The single-channel passing capacity during peak hours is generally lower than 6,000 people per hour, which is difficult to meet the needs of megacities, resulting in resource waste or safety hazards.

[0008] In view of this, an automatic ticket checking method and an automatic ticket checking device for subway tracks are specifically proposed. Summary of the Invention

[0009] Aiming at the deficiencies of the existing technology, the present invention provides an automatic ticket checking method and an automatic ticket checking device for subway tracks, which solve the problems that the existing ticket checking machines have fixed ticket checking methods, cannot adapt to real-time passenger flow changes, are difficult to meet the needs of megacities, and result in resource waste or safety hazards.

[0010] To achieve the above objectives, the present invention is realized through the following technical solutions: An automatic ticket checking method and an automatic ticket checking device for subway tracks specifically include the following steps:

[0011] S1. Feature acquisition: The ticket checking end scans the passenger's palm vein through near-infrared light to obtain the passenger's palm vein feature set, and uses a millimeter-wave radar to capture the joint displacement signals of the passenger's walking posture and extract the gait cycle energy spectrum as the gait feature set;

[0012] S2. Feature fusion: Quantify the palm vein feature set and the gait feature set, perform complementary calculation on the palm vein feature set and the gait feature set, and dynamically allocate feature fusion weights according to the calculation results to enable the judgment of the fast passage mode and the security enhancement mode;

[0013] S3. Anomaly detection: Calculate the energy entropy of the gait cycle, generate an anomaly score by combining the variance of posture stability, and issue a security linkage warning when the anomaly score exceeds the set threshold.

[0014] The present invention is further configured that: the method for quantifying the palm vein feature set and the gait feature set in the S2 includes:

[0015] H(V, G) = -∑ v∈V ∑ g∈G p(v, g) log p(v, g)

[0016] In the formula, V is the palm vein feature set, G is the gait feature set, H(V, G) is the joint entropy of the palm vein feature set and the gait feature set, v is the palm vein feature, g is the gait feature, and p(v, g) is the probability that the palm vein feature and the gait feature appear simultaneously.

[0017] The present invention is further configured that: the method for calculating the complementarity of the palm vein feature set and the gait feature in the S2 includes:

[0018] I(V, G) = H(V) + H(G) - H(V, G)

[0019] In the formula, I(V, G) is the mutual information value of the palm vein feature set and the gait feature set, H(V) is the marginal entropy of the palm vein feature set, and H(G) is the marginal entropy of the gait feature set.

[0020] The present invention is further configured that: the method for dynamically allocating feature fusion weights according to the calculation results in the S2 includes:

[0021] A1. Define the state space:

[0022] s t = [n t , τ t , c t , α t

[0023] In the formula, n t is the number of real-time passengers in the channel, τ t is the train arrival interval time, c t is the carriage crowding degree, and α t is the current fusion weight;

[0024] A2. Define the action space:

[0025]

[0026] A3. Define the reward function:

[0027] R = -0.7·E[T w-0.3·E[L r

[0028] Wherein, R is the immediate reward, T w is the average waiting time of passengers, L r is the safety risk index;

[0029] A4. Based on the Q-learning algorithm, select the action a according to the state s t for mode switching control:

[0030] When a = 0, let α t+1 = α t , and maintain the current mode;

[0031] When a = 1, let α t+1 ≥G s1 , where G s1 is the preset weight allocation threshold, and enable the fast-pass mode;

[0032] When a = 2, let α t+1 <G s1 , and enable the safety enhancement mode.

[0033] The present invention is further configured that: only the palmar vein feature is verified in the fast-pass mode;

[0034] Both the palmar vein feature and the gait feature are verified in the safety enhancement mode.

[0035] The present invention is further configured that: the method for calculating the gait cycle energy entropy in S3 includes:

[0036] H(E)=-∑ ω p(ω)logp(ω)

[0037] Wherein, H(E) is the energy entropy of the gait cycle energy spectrum, and p(ω) is the energy probability density at the frequency ω.

[0038] The present invention is further configured that: the method for generating the anomaly score by combining the attitude stability variance in S3 includes:

[0039]

[0040]

[0041] Wherein, σθ is the variance of the joint angle θ, T is the gait cycle time, S a is the anomaly score value, θ(t) is the joint angle at time t, is the average joint angle, H max is the maximum value of the energy entropy of the gait cycle energy spectrum, and σθ,max is the maximum value of the attitude variance.​

[0042] The present invention also discloses an automatic ticket checking device for subway tracks, comprising:

[0043] A wide-angle camera for collecting gait features;

[0044] A millimeter-wave radar for capturing the walking postures of passengers;

[0045] An infrared palm vein scanning module for collecting the palm vein features of passengers;

[0046] An OLED display screen for displaying the ticket checking status and passenger information.

[0047] The present invention provides an automatic ticket checking method for subway tracks and its automatic ticket checking device. It has the following

[0048] Beneficial effects:

[0049] Through the cooperation of palm vein and gait recognition, the present invention improves the security of the ticket checking system in a non-contact acquisition manner, reduces the false recognition rate, effectively improves the accuracy of detecting abnormal behaviors, and provides self-learning ability through the reinforcement learning algorithm. It can automatically adjust the ticket checking strategy according to the real-time passenger flow and operation status, realize the optimal allocation of resources for multi-modal recognition, balance efficiency and safety, so as to reduce the congestion degree during peak hours and provide passengers with a travel experience with high satisfaction. Description of the Drawings

[0050] Figure 1 It is a schematic flow chart of the present invention. Specific Embodiments

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention.

[0052] Please refer to Figure 1 , the embodiments of the present invention provide the following technical solutions: An automatic ticket checking method for subway tracks, specifically comprising the following steps:

[0053] S1. Feature acquisition: An infrared palm vein scanning module is mounted in the middle layer of the ticket checking end, with a penetration depth of 3-5 mm and a resolution of 500 dpi. The palm veins of passengers are scanned by 850 nm near-infrared light to obtain the set of palm vein features of passengers V = {v1, v2,..., v n}, where v nLet \(T_n\) be the texture feature of the \(n\)th vein. A wide-angle camera and a millimeter-wave radar are mounted on the upper layer of the ticket-checking end. The wide-angle camera is an 8-megapixel wide-angle camera with a field of view of 120°. The millimeter-wave radar, with a frequency band of 24 GHz - 77 GHz, is used to capture the joint displacement signals of the passenger's walking posture and extract the gait cycle energy spectrum as the gait feature set.

[0054] Among them, the palm vein feature has uniqueness and stability, and is difficult to forge, which can effectively improve the security of the ticket-checking system. The gait feature has the characteristics of non-contact and non-cooperation, and can complete feature acquisition during the natural walking process of passengers, improving the ticket-checking efficiency.

[0055] S2. Feature fusion: Quantify the palm vein feature set and the gait feature set. The quantization formula is:

[0056] \(H(V, G)=-\sum\) v∈V \(\sum\) g∈G \(p(v, g)\log p(v, g)\)

[0057] In the formula, \(V\) is the palm vein feature set, \(G\) is the gait feature set, \(H(V, G)\) is the joint entropy of the palm vein feature set and the gait feature set, \(v\) is the palm vein feature, \(g\) is the gait feature, and \(p(v, g)\) is the probability that the palm vein feature and the gait feature appear simultaneously;

[0058] Calculate the complementarity of the palm vein feature set and the gait feature set. The calculation formula is:

[0059] \(I(V, G)=H(V)+H(G)-H(V, G)\)

[0060] In the formula, \(I(V, G)\) is the mutual information value of the palm vein feature set and the gait feature set, \(H(V)\) is the marginal entropy of the palm vein feature set, and \(H(G)\) is the marginal entropy of the gait feature set;

[0061] According to the complementarity calculation result, dynamically allocate the feature fusion weight, and judge whether to enable the fast-pass mode and the security enhancement mode. Among them, when \(I(V, G)\geq G\) s1 where \(G\) s1 is the preset weight allocation threshold, it is judged that the palm vein feature and the gait feature are in a highly correlated state, and the feature fusion weight is allocated as: \(I(V, G)\). Enable the fast-pass mode. In the fast-pass mode, only the palm vein feature is verified to shorten the ticket-checking time;

[0062] When \(I(V, G)<G\) s1 it is judged that the palm vein feature and the gait feature are in a low correlated state, and the feature fusion weight is allocated as: \(I(V, G)\). Enable the security enhancement mode. In the security enhancement mode, the palm vein feature and the gait feature are verified.

[0063] As an optimal solution, to achieve the dynamic adjustment of the fast passage mode and the safety enhancement mode, it is implemented through reinforcement learning. Specifically:

[0064] A1. Define the state space:

[0065] s t =[n t , τ t , c t , α t

[0066] In the formula, n t is the number of real-time passengers in the channel, τ t is the train arrival interval time, c t is the carriage congestion degree, and α t is the current fusion weight;

[0067] A2. Define the action space:

[0068]

[0069] A3. Define the reward function:

[0070] R=-0.7·E[T w -0.3·E[L r

[0071] In the formula, R is the immediate reward, T w is the average passenger waiting time, and L r is the safety risk index;

[0072] A4. Update the Q-value table through the Q-learning algorithm to obtain the optimal ticket checking strategy. The update formula is:

[0073] ΔQ=η[R+γmax aa Q w -Q]

[0074] In the formula, ΔQ is the change in the Q value, η is the learning rate, preferably 0.1, γ is the discount factor, preferably 0.9, aa is the action of the next state, and max aa Q w is the maximum future Q value, and Q is the current Q value;

[0075] Based on the Q-learning algorithm, select the action a according to the state s t to perform mode switching control:

[0076] When a = 0, let α t+1 =α t , and maintain the current mode; ​​

[0077] When a = 1, let α t+1 ≥G s1 , where G s1 is a preset weight allocation threshold, preferably 0.7, and the fast passage mode is enabled;

[0078] When a = 2, let α t+1 <G s1 , and the security enhancement mode is enabled.

[0079] S3. Abnormality detection: Calculate the gait cycle energy entropy, and the calculation formula is:

[0080] H(E)=-∑ ω p(ω)logp(ω)

[0081] In the formula, H(E) is the energy entropy of the gait cycle energy spectrum, p(ω) is the energy probability density at frequency ω, and abnormal gaits usually have higher energy entropy. Therefore, the gait energy entropy can be an important indicator for judging abnormal passenger behavior;

[0082] Generate an abnormality score by combining the posture stability variance, and the calculation formula is:

[0083]

[0084] In the formula, σθ is the variance of the joint angle θ, T is the gait cycle time, S a is the abnormality score value, θ(t) is the joint angle at time t, is the average joint angle, H max is the maximum value of the gait cycle energy spectrum energy entropy, σθ,max is the maximum value of the posture variance. The posture stability variance reflects the stability of the passenger's walking posture, and abnormal behavior will cause the variance to increase, where abnormal behavior includes but is not limited to staggering and stagnation;

[0085] When S a ≥G s2 , where G s2 is a set threshold, preferably 0.6, it is determined that the passenger behavior is abnormal, triggering an anti-theft linkage warning and notifying the station staff to handle it.

[0086] Simulation experiment 1

[0087] In the morning rush hour commuting scenario.

[0088] Passenger flow characteristics: n t = 50 people / channel, τ t = 1.5 min.

[0089] The mutual information value I(V, G)=0.82>0.7, start the fast channel mode, and only verify the palm vein.

[0090] It is found that 8,800 people can pass through a single channel per hour, and the average waiting time is 4.2 s.

[0091] Simulation Experiment 2

[0092] In the low - peak scenario at night.

[0093] Passenger flow characteristics: n t = 5 people / channel, τ t = 8 min.

[0094] The mutual information value I(V, G) = 0.65 < 0.7, start the security enhancement mode, and verify the palm vein and gait characteristics.

[0095] It is found that the gait - characteristic verification intercepts 1 case of trailing attempt, and the false recognition rate is 0.

[0096] In summary, it can be found that through the mutual - information - driven dynamic weight allocation, the present invention realizes the adaptability of the ticket - checking system to complex operating environments, and achieves an effective balance among efficiency, security, and user experience.

[0097] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article or device.

[0098] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic ticket checking method for subway tracks, characterized in that, Specifically, it includes the following steps: S1. Feature acquisition: The ticket checking terminal scans the passenger's palm vein with near-infrared light to obtain the passenger's palm vein feature set, and uses a millimeter-wave radar to capture the joint displacement signal of the passenger's walking posture, and extracts the gait cycle energy spectrum as the gait feature set; S2. Feature fusion: Quantify the palm vein feature set and the gait feature set, calculate the complementarity of the palm vein feature set and the gait feature set, and dynamically allocate feature fusion weights according to the calculation results to enable the judgment of the fast passage mode and the security enhancement mode; S3. Abnormality detection: Calculate the gait cycle energy entropy, generate an abnormality score in combination with the posture stability variance, and issue a security linkage warning when the abnormality score exceeds the set threshold.

2. The automatic ticket checking method for subway tracks according to claim 1, characterized in that, The method of quantifying the palm vein feature set and the gait feature set in S2 includes: H(V, G) = -∑ v∈V ∑ g∈G p(v, g) log p(v, g) In the formula, V is the palm vein feature set, G is the gait feature set, H(V, G) is the joint entropy of the palm vein feature set and the gait feature set, v is the palm vein feature, g is the gait feature, and p(v, g) is the probability that the palm vein feature and the gait feature appear simultaneously.

3. The automatic subway ticket checking method according to claim 1, characterized in that, The method of calculating the complementarity of the palm vein feature set and the gait feature set in S2 includes: I(V, G) = H(V) + H(G) - H(V, G) In the formula, I(V, G) is the mutual information value of the palm vein feature set and the gait feature set, H(V) is the marginal entropy of the palm vein feature set, and H(G) is the marginal entropy of the gait feature set.

4. The automatic ticket checking method for subway tracks according to claim 3, characterized in that, The method of dynamically allocating feature fusion weights according to the calculation results in S2 includes: A1. Define the state space: s t = [n t , τ t , c t , α t ​ Where n t is the real-time number of passengers in the channel, τ t is the train arrival interval, c t is the car congestion degree, α t is the current fusion weight; A2. Define the action space: A3. Define the reward function: R = -0.7·E[T w -0.3·E[L r ​ Wherein, R is the immediate reward, T w is the average waiting time of passengers, L r is the safety risk index; A4. Based on the Q-learning algorithm, select action a according to state s t for mode switching control: When a = 0, let α t+1 = α t , and maintain the current mode; When a = 1, let α t+1 ≥G s1 , where G s1 is a preset weight assignment threshold, and enable the fast passage mode; When a = 2, let α t+1 <G s1 , enable the security enhancement mode.

5. The automatic ticket-checking method for subway tracks according to claim 4, wherein Only the palm vein feature is verified in the fast passage mode; Both the palm vein feature and the gait feature are verified in the security enhancement mode.

6. The automatic ticket checking method for subway tracks according to claim 1, characterized in that, The method of calculating the gait cycle energy entropy in S3 includes: H(E)= -∑ ω p(ω) log p(ω) In the formula, H(E) is the energy entropy of the gait cycle energy spectrum, and p(ω) is the energy probability density at frequency ω.

7. A subway track automatic ticket checking method according to claim 1, characterized in that, The method of generating an abnormality score in combination with the posture stability variance in S3 includes: where σθ is the variance of the joint angle θ, T is the gait cycle time, S a is the abnormal score value, θ(t) is the joint angle at time t, is the average joint angle, H max is the maximum value of the energy entropy of the gait cycle energy spectrum, and σθ,max is the maximum value of the posture variance.

8. An automatic subway track ticket checking device, applied to the automatic subway track ticket checking method according to any one of claims 1-7, characterized in that, Including: A wide-angle camera for collecting gait features; A millimeter-wave radar for capturing the passenger's walking posture; An infrared palm vein scanning module for collecting the passenger's palm vein features; An OLED display screen for displaying the ticket checking status and passenger information.

Images