Railway station contact operation monitoring method and system based on intelligent identification

Abstract

The invention relates to the technical field of operation monitoring, and particularly discloses a railway station contact operation monitoring method and system based on intelligent identification, and the method comprises the following steps: S1, selecting a station contact person sample, and collecting an action video sample when the station contact person uses communication equipment; public limb action features of all the samples are extracted through a limb analysis model; face images of the samples during task execution are collected, and the blinking frequency of each sample is calculated; generating a unified frequency reference value based on all the blinking frequencies; s2, determining a contact monitoring time length, and delimiting a contact monitoring time period based on the monitoring time length after monitoring a critical moment of train state change in real time; acquiring real-time video data of a station contact person, extracting limb action features of the station contact person through a limb analysis model, and calculating the similarity between the limb action features and the common features; collecting a face image of a parking contact person in real time, and analyzing the current blinking frequency; and S3, generating a working state index in combination with the limb index and the concentration index.

Description

Technical Field

[0001] The present invention relates to the technical field of operation monitoring, and particularly relates to a railway station liaison operation monitoring method and system based on intelligent recognition. Background Art

[0002] Railway station liaison operation monitoring is an important measure taken during railway construction and maintenance operations to ensure the safety of operating personnel and guarantee the order of railway transportation; through various monitoring methods and strict monitoring processes, it ensures the safety of operating personnel and the smoothness of railway transportation.

[0003] In railway liaison operations, since all links require staff to follow up carefully, it is necessary for staff to maintain a high degree of concentration. However, existing monitoring technologies mainly focus on the working status of on-site personnel, but the monitoring of the working status of station liaison officers is insufficient. This may not only lead to a decline in work efficiency, but also may cause safety accidents due to inattentiveness. Therefore, it is particularly important to develop a technology that can effectively monitor the working status of station liaison officers. Summary of the Invention

[0004] The purpose of the present invention is to provide a railway station liaison operation monitoring method and system based on intelligent recognition to solve the above technical problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A railway station liaison operation monitoring method based on intelligent recognition includes the following steps: Step S1: Select a number of station liaison officer samples, obtain action video samples of the station liaison officer samples when using communication devices for liaison, and establish a body analysis model. Through the body analysis model, obtain the common body action characteristics of the station liaison officer samples in all action video samples; Set a work task, obtain facial image samples of each station liaison officer sample when performing the work task, obtain the blink frequency of each station liaison officer sample according to the facial image samples, and obtain a frequency reference value according to all blink frequencies; Step S2: Obtain the duration of all action video samples, and set a liaison monitoring duration according to the duration; Real-time obtain the status of the train, where the status includes parking and running; Record the moment when the status of the train changes as the change moment, and select a liaison monitoring time period after the change moment according to the liaison monitoring duration; Obtain the video data of the station liaison officer during the liaison monitoring time period, input the video data into the body analysis model, obtain the body action characteristics of the station liaison officer in the video data; and obtain the similarity between the body action characteristics and the common body action characteristics, Obtain the workbench of the on-site liaison officer, set monitoring points on the workbench, and the monitoring points are used to obtain the facial images of the on-site liaison officer in real time; obtain the blinking frequency of the on-site liaison officer according to the facial images; Step S3: Set the limb index I 1 , if the similarity is greater than or equal to the preset similarity threshold, record the limb index I 1 = 1, otherwise, record the limb index I 1 = 0; set the concentration index I 2 , if the blinking frequency exceeds the frequency reference value, record the concentration index I 2 = 1, otherwise, record the concentration index I 2 = 0; obtain the working status index according to the concentration index and the limb index. When and only when the working status index WI = 1, the working status of the on-site liaison officer is normal.

[0006] As a further solution of the present invention: The acquisition of the action video sample is based on an action capture device, and the action video sample includes action videos of the on-site liaison officer sample using a communication device obtained at different angles.

[0007] As a further solution of the present invention: The acquisition process of the blinking frequency of the on-site liaison officer sample includes: Establish an image recognition model, number all facial image samples in time series, and perform image recognition on the facial image samples through the image recognition model to obtain the recognition result, and the recognition result includes closing eyes and opening eyes; obtain the numbers {n 1 , n 2 ,..., n m} of the facial image samples with the recognition result of closing eyes, where n m represents the number of the m-th facial image sample with the recognition result of closing eyes, and m is the total number of facial image samples with the recognition result of closing eyes; obtain the blinking frequency , where n i represents the number of the i-th facial image sample with the recognition result of closing eyes, i ∈ [1, m - 1] and i is a positive integer, and t is the unit time.

[0008] As a further solution of the present invention: The liaison monitoring duration is the average value of the durations of all action video samples.

[0009] As a further solution of the present invention: When the working status index WI = 0, the working status of the on-site liaison officer is abnormal.

[0010] As a further solution of the present invention: The establishment process of the limb analysis model includes: Divide the action video sample into several image frames, label and segment the limb parts of the on-site liaison officer sample in the image frames based on computer vision technology, mark the joint points and the outlines of the limbs, record the joint points and the outlines as limb information, and mark the action names corresponding to the limb information to obtain sample data; establish an initial model based on a convolutional neural network, and train the initial model with the sample data to obtain a limb analysis model.

[0011] As a further solution of the present invention: the process of obtaining the frequency reference value includes: Obtain the average value of all blinking frequencies, denoted as Bf ave , and obtain the standard deviation s of all blinking frequencies, then the frequency reference value Frv = Bf ave -z×s, where z is a multiple threshold, and z ∈ [0, 3].

[0012] A railway on-site liaison operation monitoring system based on intelligent recognition includes: Sample acquisition module: Select several on-site liaison officer samples, obtain the action video samples of the on-site liaison officer samples when using communication equipment for liaison, and establish a limb analysis model. Through the limb analysis model, obtain the common limb action characteristics of the on-site liaison officer samples in all action video samples; Set the work tasks, obtain the facial image samples of each on-site liaison officer sample when performing the work tasks, obtain the blinking frequencies of each on-site liaison officer sample according to the facial image samples, and obtain the frequency reference value according to all blinking frequencies; Monitoring module: Obtain the duration of all action video samples, and set the liaison monitoring duration according to the duration; obtain the status of the train in real time, and the status includes parking and running; record the moment when the status of the train changes as the change moment, and select the liaison monitoring time period after the change moment according to the liaison monitoring duration; Obtain the video data of the on-site liaison officer during the liaison monitoring time period, input the video data into the limb analysis model, and obtain the limb action characteristics of the on-site liaison officer in the video data; and obtain the similarity between the limb action characteristics and the common limb action characteristics, Obtain the workbench of the on-site liaison officer, set monitoring points on the workbench, and the monitoring points are used to obtain the facial images of the on-site liaison officer in real time; obtain the blinking frequency of the on-site liaison officer according to the facial images; Judgment module: Set the limb index I 1 , if the similarity is greater than or equal to the preset similarity threshold, then record the limb index I 1 = 1, otherwise, record the limb index I 1 = 0; set the concentration index I2 If the blink frequency exceeds the frequency reference value, record the concentration index I 2 = 1; otherwise, record the concentration index I 2 = 0; According to the concentration index and the body index, obtain the working status index. When and only when the working status index WI = 1, the working status of the on-site liaison officer is normal.

[0013] Advantages of the present invention: By establishing a body analysis model and a blink frequency reference value, the present invention can more accurately evaluate the working status of on-site liaison officers, thereby reducing false alarms and missed reports; this method can obtain train status and video data of on-site liaison officers in real time, and make judgments and adjustments on the working status in a timely manner, which helps to improve the response speed and work efficiency; by automatically analyzing video data and facial images through intelligent algorithms, the need for manual intervention is reduced, and the automation level of the monitoring process is improved; by continuously monitoring the working status of on-site liaison officers, irregular actions or inattentive states can be detected and corrected in a timely manner, which helps to prevent accidents and ensure the safety of railway transportation; the collected data can be used for further analysis and optimization of work processes, providing decision-making support for management and promoting the improvement and optimization of operation processes. Description of the Drawings

[0014] The present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 It is a schematic flow chart of a method for monitoring railway on-site liaison operations based on intelligent recognition according to the present invention. Detailed Embodiments

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0017] Please refer to Figure 1 as shown, the present invention is a method for monitoring railway on-site liaison operations based on intelligent recognition, including the following steps: Step S1: Select a number of on-site liaison officer samples, obtain action video samples of the on-site liaison officer samples when using communication equipment for liaison, and establish a body analysis model. Through the body analysis model, obtain the common body action characteristics of the on-site liaison officer samples in all action video samples; Set a work task, obtain facial image samples of each on-site liaison officer when performing the work task, obtain the blinking frequency of each on-site liaison officer sample based on the facial image samples, and obtain a frequency reference value based on all the blinking frequencies; It can be understood that the core of the above steps is to provide action specifications and physiological benchmarks for subsequent monitoring through data-driven standardized modeling; select multiple groups of on-site liaison officer samples, record action video samples when they use communication devices, and use computer vision technologies (such as pose estimation, action segmentation) to analyze the action video samples and extract key action nodes (such as gestures, body postures, device operation procedures); extract common action patterns from multiple samples through machine learning (such as clustering analysis, feature dimensionality reduction) to form a standardized public limb action feature library; eliminate individual differences and establish a standard action template for subsequent compliance comparison of real-time actions; At the same time, when the on-site liaison officer performs a specific task, collect facial image samples through a camera, use an image recognition algorithm (such as Haar cascade, deep learning eye segmentation) to detect the opening and closing state of the eyes, and count the time consumed for a single blink; conduct statistical analysis on the blinking frequencies of all samples to generate a frequency reference value; It should be noted that by combining action and physiological indicators (limbs + blinking), the limitations of single indicators are avoided; establish a benchmark through sample data, adapt to different scenarios and individual differences, improve the robustness of the monitoring system, transform artificial experience into a data model, and achieve an upgrade from subjective judgment to objective intelligent analysis; In a preferred embodiment of the present invention, the acquisition of the action video samples is based on an action capture device, and the action video samples include action videos of on-site liaison officer samples using communication devices obtained from different angles; It can be understood that a plurality of cameras or sensors are arranged around the on-site liaison officer to synchronously collect action videos from different angles, and the data is integrated through a multi-view fusion algorithm (such as 3D reconstruction, feature matching) to generate a complete action model; ensure the comprehensiveness of the action data, avoid the lack of action information caused by a single perspective, improve the robustness of action analysis, and adapt to different scenarios and posture changes; In a preferred embodiment of the present invention, the establishment process of the limb analysis model includes: Divide the action video samples into several image frames, label and segment the limb parts of the on-site liaison officer samples in the image frames based on computer vision technology, mark the joint points and the outlines of the limbs, record the joint points and the outlines as limb information, and mark the action names corresponding to the limb information to obtain sample data; establish an initial model based on a convolutional neural network, and train the initial model through the sample data to obtain a limb analysis model; It should be noted that computer vision technology is used to label and segment limb parts in image frames, mark joint points (such as elbows, knees) and limb contours (such as arms, legs) to form structured limb information; label each action with a corresponding action name (such as raising a hand, turning around) to generate labeled sample data; In a preferred embodiment of the present invention, the process of obtaining the blink frequency of the on-site liaison officer sample includes: Establish an image recognition model, number all facial image samples in time series, and perform image recognition on the facial image samples through the image recognition model to obtain recognition results, where the recognition results include closed eyes and open eyes; obtain the numbers {n 1 ,n 2 ,...,n m} of the facial image samples with the recognition result of closed eyes, where n m represents the number of the mth facial image sample with the recognition result of closed eyes, and m is the total number of facial image samples with the recognition result of closed eyes; obtain the blink frequency where n i represents the number of the ith facial image sample with the recognition result of closed eyes, i ∈ [1, m - 1] and i is a positive integer, and t is the unit time; It should be noted that calculate the time interval between adjacent closed-eye images. For example, if the number of the ith closed-eye image is n i and the number of the (i + 1)th closed-eye image is n i+1 , then the time interval between them is n i+1 -n i , which represents the total time span from the first closed eye to the last closed eye; m - 1 represents the number of times of closing eyes minus 1 (i.e., the number of blinks); the unit time t is used to standardize the frequency value (such as the number of blinks per minute); It is worth noting that the blink frequency is an important physiological indicator to measure concentration, because it can directly reflect a person's attention state and fatigue level. When a person is in a daze or sleepy, the blink frequency is usually significantly lower than the frequency reference value; when in a daze, the person's brain is in a relaxed or unconscious state, and the muscle activity of the eyes decreases, resulting in a decrease in the blink frequency. At this time, when the attention is scattered or there is a lack of external stimuli, the brain's control over eye movement weakens, and the blink frequency naturally decreases; when sleepy, fatigue causes the tension of the eye muscles to decrease, the closing time of the eyelids to prolong, and thus the number of blinks to decrease. At this time, as the fatigue increases, the brain's alertness decreases, and the ability to regulate eye movement weakens, resulting in a decrease in the blink frequency; In a preferred embodiment of the present invention, the process of obtaining the frequency reference value includes: Obtain the average value of all blink frequencies, denoted as Bf aveand obtain the standard deviation s of all the blinking frequencies, then the frequency reference value Frv = Bf ave - z×s, where z is a multiple threshold, and z ∈ [0, 3]; It can be understood that there are differences in the blinking frequency baselines of different people (such as those with naturally high or low blinking frequencies). By combining the average value and the standard deviation with the multiple threshold, the reference value is dynamically adjusted to better fit the actual physiological characteristics of individuals or groups; avoiding misjudgment caused by individual differences and improving the accuracy and fairness of the monitoring system; by adjusting the multiple threshold, such as the smaller the z value, the closer the reference value is to the average value and the stricter the monitoring; the larger the z value, the looser the reference value and the more individual differences are accommodated; Step S2: Obtain the duration of all the action video samples, set the contact monitoring duration according to the duration; obtain the status of the train in real time, and the status includes stopping and running; record the moment when the status of the train changes as the change moment, and select a contact monitoring time period after the change moment according to the contact monitoring duration; Obtain the video data of the on - site liaison officer during the contact monitoring time period, input the video data into the limb analysis model, and obtain the limb movement characteristics of the on - site liaison officer in the video data; and obtain the similarity between the limb movement characteristics and the public limb movement characteristics, Obtain the workbench of the on - site liaison officer, set monitoring points on the workbench, and the monitoring points are used to obtain the facial images of the on - site liaison officer in real time; obtain the blinking frequency of the on - site liaison officer according to the facial images; It should be noted that the number of the contact time periods is not one, but several within the entire operation time period. The moment of the train status change is an instant, which is the moment when the train changes from static to moving or from moving to static, and the contact time period is a period of time before or after the change moment to monitor whether the on - site liaison officer has the behavior of timely contact during the contact time period; It can be understood that based on the duration of all the action video samples, a reasonable contact monitoring duration is determined; according to the critical moment of the train status change (such as stopping and then starting), the contact monitoring time period is delimited to ensure monitoring during the critical period; avoiding the waste of resources for all - weather monitoring, focusing on high - risk periods and improving the monitoring efficiency; the train status is divided into two types: stopping and running, and the moment of status change (such as stopping and then starting) is marked as the change moment; after the change moment, according to the set contact monitoring duration, the monitoring task is started; ensuring real - time monitoring of the working status of the on - site liaison officer during the critical period of the train status change (such as starting or stopping); by evaluating whether the on - site liaison officer has the action of using communication equipment at the moment of the train status change; As a preferred embodiment of the present invention, the contact monitoring duration is the average value of the durations of all the action video samples; Step S3: Set the limb index I 1 , if the similarity is greater than or equal to a preset similarity threshold, record the limb index I 1 = 1, otherwise, record the limb index I 1 = 0; Set the concentration index I 2 , if the blink frequency exceeds the frequency reference value, record the concentration index I 2 = 1, otherwise, record the concentration index I 2 = 0; Obtain the working status index according to the concentration index and the limb index. When and only when the working status index WI = 1, the working status of the on-site liaison officer is normal; It should be noted that the limb index is used to evaluate the action standardization of the on-site liaison officer; if the similarity between the real-time action characteristics and the public limb action characteristics ≥ the preset threshold, then I 1 = 1 (normal action); otherwise I 2 = 0 (abnormal action); Calculate the similarity between the real-time action characteristics and the standard action template through the limb analysis model. The similarity threshold is determined according to historical data or experiments to ensure the accuracy of the judgment; Quantify the action compliance to provide a basis for the working status evaluation; It should be noted that the concentration index is used to evaluate the concentration of the on-site liaison officer. If the real-time blink frequency ≤ the frequency reference value, then I 2 = 1 (concentrated); otherwise I 2 = 0 (distracted or fatigued); Calculate the blink frequency through image recognition technology and compare it with the frequency reference value. The frequency reference value is dynamically adjusted based on population data to adapt to individual differences; Quantify the concentration to provide a supplementary index for the working status evaluation; As a preferred embodiment of the present invention, when the working status index WI = 0, the working status of the on-site liaison officer is abnormal; It should be noted that the working status index is the logical AND result of the limb index and the concentration index. Integrate the limb index and the concentration index through logical operations to generate a comprehensive evaluation result to support real-time monitoring and dynamic early warning; Comprehensively evaluate the working status through multi-dimensional indicators (action + concentration) to improve the accuracy and reliability of the judgment.

[0018] It is worth noting that the present invention also includes an intelligent judgment of the on-site situation in the station. The process of the intelligent judgment includes: 1. Intelligent recognition and notification of train arrival information: Train arrival detection: Real-time obtain the train position and running status through sensors, track circuits or train positioning systems; Use artificial intelligence algorithms to predict the train arrival time to ensure early warning; Automatic notification: When a train is detected approaching, an automatic notification is sent to the on-site liaison officer at the station; the on-site liaison officer at the station shall respond in a timely manner and notify the on-site operation personnel through communication equipment; Action confirmation: Through video monitoring or motion capture technology, it is monitored in real time whether the on-site liaison officer at the station has performed the notification action; if there is no timely response, a warning is triggered to remind the liaison officer to fulfill their duties; 2. Intelligent verification of personnel getting off the track: Satellite positioning and video monitoring: Through the satellite positioning equipment worn by on-site operation personnel, their location information is obtained in real time; combined with on-site video monitoring, it is confirmed whether the personnel get off the track and enter the safe area as required; Distance calculation and safety judgment: Calculate the real-time distance between the personnel's position and the track, and judge whether it meets the safety standards; use geographic information system (GIS) and spatial analysis algorithms to dynamically delimit the boundary of the safe area to ensure the accuracy of the judgment; Intelligent verification and warning: If the personnel do not get off the track or enter the safe area as required, a warning is automatically triggered to notify the on-site liaison officer at the station and on-site management personnel; combined with historical data and machine learning models, optimize the delimitation and judgment logic of the safe area to improve the intelligence level of the system; 3. Data integration and comprehensive analysis: Multi-source data fusion: Integrate multi-source data such as train position, personnel positioning, and video monitoring to construct a comprehensive on-site operation situation map; through big data analysis and visualization technology, display the status of trains, personnel, and equipment in real time to assist decision-making; Abnormality detection and handling: Use abnormality detection algorithms to identify behaviors that do not meet safety specifications; automatically generate handling suggestions and send them to relevant personnel through the system; An intelligent recognition-based railway on-site liaison operation monitoring system, including: Sample collection module: Select a number of on-site liaison officer samples, obtain action video samples of the on-site liaison officer samples when using communication equipment for liaison, and establish a limb analysis model. Through the limb analysis model, obtain the common limb action characteristics of the on-site liaison officer samples in all action video samples; Set work tasks, obtain facial image samples of each on-site liaison officer sample when performing the work tasks, obtain the blink frequency of each on-site liaison officer sample according to the facial image samples, and obtain the frequency reference value according to all blink frequencies; Monitoring module: Obtain the duration of all action video samples, and set the liaison monitoring duration according to the duration; obtain the status of the train in real time, and the status includes parking and driving; record the moment when the status of the train changes as the change moment, and select the liaison monitoring time period after the change moment according to the liaison monitoring duration; Obtain the video data of the on-site liaison officer during the liaison monitoring time period, input the video data into the limb analysis model, and obtain the limb movement characteristics of the on-site liaison officer in the video data; and obtain the similarity between the limb movement characteristics and the public limb movement characteristics. Obtain the workbench of the on-site liaison officer, set monitoring points on the workbench, and the monitoring points are used to obtain the facial images of the on-site liaison officer in real time; obtain the blink frequency of the on-site liaison officer according to the facial images. Judgment module: Set the limb index I 1 , if the similarity is greater than or equal to the preset similarity threshold, record the limb index I 1 = 1, otherwise, record the limb index I 1 = 0; Set the concentration index I 2 , if the blink frequency exceeds the frequency reference value, record the concentration index I 2 = 1, otherwise, record the concentration index I 2 = 0; Obtain the work status index according to the concentration index and the limb index. When and only when the work status index WI = 1, the work status of the on-site liaison officer is normal.

[0019] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equal changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. A railway station liaison operation monitoring method based on intelligent identification, characterized in that: The following steps are involved: Step S1: Select several samples of on-site liaison officers, obtain action video samples of the samples of on-site liaison officers when using communication equipment to make contacts, and establish a body analysis model. Through the body analysis model, obtain the common body movement features of the samples of on-site liaison officers in all action video samples; Setting a work task, obtaining facial image samples of each on-site liaison officer sample when performing the work task, obtaining the blinking frequency of each on-site liaison officer sample according to the facial image samples, and obtaining a frequency reference value according to all blinking frequencies; Step S2: Obtain the duration of all action video samples, and set the contact monitoring duration according to the duration; obtain the state of the train in real time, the state including parking and driving; record the time when the state of the train changes as the change time, and select the contact monitoring time period after the change time according to the contact monitoring duration; Obtain video data of the station liaison officer during the contact monitoring period, input the video data into the body analysis model, obtain body movement features of the station liaison officer in the video data; and obtain the similarity between the body movement features and the public body movement features, Obtaining a workbench of the station liaison officer, setting a monitoring point on the workbench, wherein the monitoring point is used to obtain a facial image of the station liaison officer in real time; obtaining a blinking frequency of the station liaison officer according to the facial image; Step S3: Set the limb index I1. If the similarity is greater than or equal to a preset similarity threshold, record the limb index I1=1, otherwise, record the limb index I1=0; set the concentration index I2. If the blinking frequency exceeds the frequency reference value, record the concentration index I2=1, otherwise, record the concentration index I2=0; obtain the work status index based on the concentration index and the limb index. If and only if the work status index WI=1, there is no abnormality in the work status of the on-site liaison officer.

2. According to the method for monitoring railway station liaison operations based on intelligent identification according to claim 1, it is characterized in that: In step S1, the acquisition of the action video samples is based on a motion capture device, and the action video samples include action videos of the on-site liaison officer samples using communication equipment acquired at different angles.

3. The method for monitoring railway station liaison operations based on intelligent identification according to claim 1 is characterized in that: In step S1, the process of establishing the limb analysis model includes: The action video samples are divided into a number of image frames, and the limb parts of the resident liaison officer samples in the image frames are annotated and segmented based on computer vision technology, the joints and the contours of the limbs are marked, the joints and the contours are recorded as limb information, and the action names corresponding to the limb information are marked to obtain sample data; an initial model is established based on a convolutional neural network, and the initial model is trained with the sample data to obtain a limb analysis model.

4. The method for monitoring railway station liaison operations based on intelligent identification according to claim 1 is characterized in that: In step S1, the process of acquiring the blinking frequency of the on-site liaison officer sample includes: Establish an image recognition model to identify all facial image samples, and perform image recognition on the facial image samples according to the time series, and obtain recognition results, wherein the recognition results include closed eyes and open eyes; obtain the numbers {n1, n2, ..., n} of the facial image samples whose recognition results are closed eyes. m }, where n m represents the number of the mth facial image sample with closed eyes as the recognition result, and m is the total number of facial image samples with closed eyes as the recognition result; the blink frequency is obtained , where n i It represents the number of the facial image sample whose recognition result is closed eyes, i∈[1,m-1] and i is a positive integer, and t is the unit time.

5. The method for monitoring railway station liaison operations based on intelligent identification according to claim 1 is characterized in that: In step S1, the process of obtaining the frequency reference value includes: Get the average of all blink frequencies, recorded as Bf ave , and obtain the standard deviation s of all blink frequencies, then the frequency reference value Frv=Bf ave -z×s, where z is the multiple threshold and z∈[0,3].

6. The method for monitoring railway station liaison operations based on intelligent identification according to claim 1, characterized in that: In step S2, the contact monitoring duration is the average duration of all action video samples.

7. The method for monitoring railway station liaison operations based on intelligent identification according to claim 1, characterized in that: In step S3, when the work status index WI=0, the work status of the on-site liaison officer is abnormal.

8. A railway station liaison operation monitoring system based on intelligent identification, characterized in that: include: Sample collection module: select a number of on-site liaison samples, obtain action video samples of the on-site liaison samples when using communication equipment to make contacts, and establish a body analysis model. Through the body analysis model, obtain the common body movement features of the on-site liaison samples in all action video samples; Setting a work task, obtaining facial image samples of each on-site liaison officer sample when performing the work task, obtaining the blinking frequency of each on-site liaison officer sample according to the facial image samples, and obtaining a frequency reference value according to all blinking frequencies; Monitoring module: obtaining the duration of all action video samples, and setting the contact monitoring duration according to the duration; obtaining the status of the train in real time, the status including parking and driving; recording the time when the status of the train changes as the change time, and selecting the contact monitoring time period after the change time according to the contact monitoring duration; Obtain video data of the station liaison officer during the contact monitoring period, input the video data into the body analysis model, obtain body movement features of the station liaison officer in the video data; and obtain the similarity between the body movement features and the public body movement features, Obtaining a workbench of the station liaison officer, setting a monitoring point on the workbench, wherein the monitoring point is used to obtain a facial image of the station liaison officer in real time; obtaining a blinking frequency of the station liaison officer according to the facial image; Judgment module: set the limb index I1, if the similarity is greater than or equal to the preset similarity threshold, then record the limb index I1=1, otherwise, record the limb index I1=0; set the concentration index I2, if the blinking frequency exceeds the frequency reference value, then record the concentration index I2=1, otherwise, record the concentration index I2=0; obtain the work status index based on the concentration index and the limb index, if and only if the work status index WI=1, the work status of the on-site liaison officer is normal.

Images