CN116229305B - Real-time evaluation method for subway station streamline congestion state based on video identification - Google Patents
Real-time evaluation method for subway station streamline congestion state based on video identification Download PDFInfo
- Publication number
- CN116229305B CN116229305B CN202211620815.0A CN202211620815A CN116229305B CN 116229305 B CN116229305 B CN 116229305B CN 202211620815 A CN202211620815 A CN 202211620815A CN 116229305 B CN116229305 B CN 116229305B
- Authority
- CN
- China
- Prior art keywords
- streamline
- flow line
- congestion
- facilities
- passenger flow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 65
- 238000012544 monitoring process Methods 0.000 claims abstract description 25
- 238000005516 engineering process Methods 0.000 claims abstract description 18
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 238000000034 method Methods 0.000 claims abstract description 11
- 239000000470 constituent Substances 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 16
- 238000009499 grossing Methods 0.000 claims description 15
- 210000002569 neuron Anatomy 0.000 claims description 7
- 238000012546 transfer Methods 0.000 claims description 7
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000011144 upstream manufacturing Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 2
- 230000008520 organization Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 abstract 1
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a real-time evaluation method for a flow line congestion state of a subway station based on video data, and belongs to the technical field of urban rail transit intelligent passenger transport management. Comprising the following steps: selecting a main passenger flow line of a subway station in a peak period as an evaluation flow line, and dividing the constituent facilities on the flow line into m service facilities and n channel facilities; the method comprises the steps of extracting passenger flow parameters of service facilities and channel facilities on a flow line by utilizing a video identification technology, calculating passenger flow line congestion state evaluation index values, inputting a probabilistic neural network model to obtain N number of flow line congestion state evaluation grade values Y j, carrying out weighted summation on Y j by taking the occurrence frequency of the N number of flow line congestion state evaluation grade values as weights, and obtaining and publishing the flow line comprehensive congestion grade value P. The method for establishing the passenger flow line congestion state evaluation index based on video identification can provide effective basis for monitoring and early warning and emergency control of the subway station dynamic or sudden large passenger flow from single-point and streamline layers, and has important effects on improving the organization efficiency and operation safety of the large passenger flow of the station.
Description
Technical Field
The invention belongs to the technical field of urban rail transit intelligent passenger transport management, and particularly relates to a real-time evaluation method for a streamline congestion state of a subway station based on video data.
Background
Along with the expansion of the urban rail transit network scale, the subway station passenger flow scale is continuously increased, and in addition, partial station early-stage facility design standards are not suitable for actual passenger flow demands, passenger flow congestion frequently occurs at partial station facilities, so that the high-density crowd conflict and congestion trampling risks are greatly increased.
The real-time accurate evaluation of the crowded state of the passenger flow at the facilities is the basis for developing the flow control of the large passenger flow of the station. The prior related technology is used for evaluating the state of the passenger flow congestion of single-point facilities or local regional facilities of a station, lacks complex characteristics of multiple concurrency, relevance transmissibility and the like of various service facilities (import and export, gate, escalator and the like) and passenger flow congestion of channel facilities on a streamline from the aspect of an access station or transfer streamline, greatly reduces the practical value of an evaluation result (the problems of failure in consideration of management and control, lag in management and control and the like of large passenger flow management and control measures formulated based on the evaluation result) and is difficult to provide scientific basis for the formulation and optimization of dynamic or sudden large passenger flow emergency treatment schemes of the station.
Disclosure of Invention
Aiming at the defects of the prior art, the system and the method collect passenger flow parameters of various facilities on the passenger flow based on the video identification technology, develop real-time evaluation of passenger flow congestion state from the streamline layer, and provide scientific basis for monitoring and early warning and emergency management and control of subway station dynamic or sudden large passenger flow.
In order to achieve the above purpose, the invention provides a real-time evaluation method for the flow line congestion state of a subway station based on video data, which comprises the following steps:
(1) Selecting a main passenger flow line of a subway station in a peak period as an evaluation flow line, and dividing the constituent facilities on the flow line into m service facilities and n channel facilities according to the traffic flow characteristics of pedestrians;
(2) Setting a streamline congestion evaluation interval as C minutes and a streamline congestion release interval as NC minutes;
(3) Extracting passenger flow parameters of the service facilities and the channel facilities on the flow line every C minutes by utilizing a video identification technology;
The passenger flow parameters extracted by the service facility comprise the passenger flow arrival rate lambda i of the section upstream of the service facility, the actual queuing captain l i and the actual queuing number r i of pedestrians in front of the service facility; the passenger flow parameters extracted by the channel facilities are pedestrian running speeds v g of 1 or more video monitoring points distributed on each channel facility;
For the service facilities, arranging video monitoring points in an allowable queuing space area, and calibrating the saturation flow rate mu i of the ith service facility in advance, namely the maximum number of people passing through per unit time, the maximum allowable queuing length L i and the queuing number R i; the passenger flow arrival rate lambda i of the ith service facility, namely the number of people actually arriving in unit time, the pedestrian actual queuing captain l i and the actual queuing number r i are acquired in real time through a video identification technology, wherein i=1, 2, 3..m;
For the channel facility, P video monitoring points are distributed in key sections such as an inlet and an outlet of the channel, a turning area, a capacity bottleneck area and the like, and the pedestrian running speed of all the monitoring points is v g, wherein: g=1, 2,3 … P, P is not less than n;
(4) Calculating a passenger flow line congestion state evaluation index value according to the passenger flow parameters extracted by each video monitoring point, wherein the passenger flow line congestion state evaluation index value comprises the total average delay time T of the service facility, the occupancy rate eta of the queuing space of the whole flow line and the pedestrian travelling speed variation coefficient C v of the channel facility on the flow line;
(4.1) determining a distribution of arrival rates lambda i of the passengers at peak time by survey data fitting, wherein the maximum departure rate is the saturation flow rate mu i of the facility, queuing delay is generated when lambda i is larger than mu i, and the area enclosed by the accumulated arrival and departure curves is the total queuing delay time T i of the passengers at the ith service facility:
The average queuing delay time t i:
Wherein: q 1 is the number of accumulated passengers arriving, and lambda i is obtained by video recognition technology according to the calculation of the distribution function of the passenger arrival rate lambda i; q 2 is the cumulative number of passengers leaving, which is the product of the saturation flow rate μ i of the ith service facility and the duration t; t is queuing duration;
The total average delay time T of various service facilities on the streamline is as follows:
(4.2) for a service facility, the full-flow line queuing space occupancy η calculation formula is:
Wherein: gamma i is the occupancy weight of queuing space of different facilities, which is obtained by expert scoring, The product of the maximum allowed queuing length L i and the number of queues R i is the maximum queuing space area allowed by station operation safety management; the actual queuing length l i and the actual queuing number r i of the pedestrians are obtained according to a video recognition technology;
(4.3) aiming at the channel facilities, representing the crowding degree of the channel facilities on the streamline by using the obtained speed fluctuation acquired by pedestrians at p channel video monitoring points, wherein the calculation formula of the channel pedestrian running speed variation coefficient C v is as follows:
wherein: the average value of the pedestrian walking speeds of p video monitoring points of the channel facility is obtained;
(5) Constructing a probabilistic neural network model PNN, taking the passenger flow streamline congestion state evaluation index value as the input of the model, evaluating the streamline congestion state in real time, and outputting N streamline congestion state evaluation grade values Y j in the streamline congestion release interval;
(6) And based on N streamline congestion state evaluation grade values Y j in the streamline congestion release interval, weighting and summing Y j by taking the occurrence frequency of each streamline congestion state evaluation grade value as a weight, obtaining a streamline comprehensive congestion grade value P in the streamline congestion release interval, and releasing.
Further, the main passenger flow line is an inbound, outbound or transfer line with large passenger flow and 1 or more frequent congestion points.
Further, the service facilities comprise an entrance, a security check machine, a gate and an escalator entrance; the channel facilities comprise a long channel for entering and exiting the station and a long channel for transferring.
Further, the probabilistic neural network model training process in the step 5 is as follows:
(5.1) constructing a training sample data set, wherein the sample value is represented by x= (C v, T, η), the sample size is Z, and the sample size is transferred to the pattern layer, and the neuron number is Z;
(5.2) performing dimensionless processing on the passenger flow line congestion state evaluation index value in the sample data set;
(5.3) initializing a smoothing factor, and adopting the same smoothing factor for samples belonging to the same category hidden layer;
(5.4) training the probabilistic neural network model, endowing the probabilistic neural network model with the smoothing factor to start training, and calculating a mode layer output, an output layer output and an error rate;
(5.5) modifying the smoothing factor and the training threshold;
(5.6) judging the error rate and the given threshold value, outputting a congestion level if the error rate is smaller than the given threshold value, entering training of the next sample if the error exceeds 1.03 times of the previous error, otherwise, adjusting a smoothing factor;
(5.7) continuing training the network until the condition is met, inputting a sample to be tested, and obtaining a streamline congestion level of the sample to be tested, wherein the streamline congestion level is divided into M levels, namely, the congestion state evaluation level value Y j = 1,2,3. j=1, 2,3.
Further, the calculation formula of the flow line comprehensive congestion level value P in the step 6 is as follows:
wherein: y j = 1,2, 3..m, j = 1,2, 3..n, Refers to the frequency with which each Y j occurs within the streamline congestion issue interval.
The invention has the beneficial effects that:
The invention considers the passenger flow congestion characteristics of various service facilities and channel facilities on the incoming and outgoing or transfer streamlines (queuing delay characteristics under the congestion state of node facilities and pedestrian speed change characteristics under the congestion state of channel facilities) and the multiple concurrency and associated transmissibility of various passenger flow congestion points on the streamlines, takes the feasibility of the video recognition technology into consideration to collect the passenger flow parameters on the main passenger flow lines, establishes the evaluation index and the evaluation method of the passenger flow streamline congestion state, can provide effective basis for monitoring and early warning and emergency management of the dynamic or sudden large passenger flow of the subway station from the single point and streamline level, and has important roles of improving the emergency organization efficiency of the dynamic or sudden large passenger flow of the subway station and guaranteeing the operation safety of the large passenger flow of the station.
Drawings
Fig. 1 is a flowchart of an overall method for evaluating a flow line congestion state of a subway station in real time based on video data according to an embodiment of the invention.
Fig. 2 is a schematic diagram of the cumulative arrival-departure of passengers according to an embodiment of the present invention.
Fig. 3 is a diagram of a probabilistic neural network according to an embodiment of the present invention.
Fig. 4 shows a subway station arrival flow chart and monitoring points according to an embodiment of the invention.
FIG. 5 illustrates the level of integrated congestion for a flow line in accordance with an embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples.
As shown in fig. 1, the embodiment of the invention provides a real-time evaluation method for a flow line congestion state of a subway station based on video data, which comprises the following steps:
S101, selecting a main passenger flow streamline of a subway station in a peak period as an evaluation streamline, and dividing the constituent facilities on the streamline into m service facilities and n channel facilities according to pedestrian traffic characteristics;
The primary traffic flow line is an inbound, outbound or transfer flow line having a large traffic flow and having 1 or more frequent congestion points.
Based on pedestrian traffic characteristics, the main passenger flow line is divided into two types of facilities, namely a service facility and a channel facility, and because the platform is a converging end point or a diverging start point of a plurality of passenger flow lines, the congestion evaluation range of a single flow line is not included, and the individual congestion state evaluation is required.
For service facilities such as entrances and exits, security check machines, gate machines, building entrances and exits, passengers can quickly pass through the facilities after arriving at the facilities under the non-congestion state, and continuous queuing is generated under the congestion state.
For the long-channel facilities such as the in-out station and the transfer long-channel, the long-channel facilities are generally in long and narrow linear structures, the pedestrian passing time is long, the pedestrian running speed is stable in the non-crowded state, and the speed fluctuation is large due to the fact that the pedestrian stops and moves in the crowded state.
S102, setting a streamline congestion evaluation interval as C minutes and a streamline congestion release interval as NC minutes;
Setting the streamline congestion evaluation interval as C minutes, wherein the lower limit value of the streamline congestion evaluation interval depends on the operation time required by a passenger flow parameter acquisition technology and a streamline congestion state evaluation algorithm, and for example, a video recognition technology can be used for obtaining 1-time streamline congestion state evaluation value in 1 minute; the flow line congestion release interval is NC minutes, N is a positive integer, the value of the positive integer depends on the actual requirement and operation feasibility of the large passenger flow control of the station, for example, the passenger flow time-space fluctuation degree in the peak period is small, and the N value is larger; the preparation time required for starting and adjusting the large passenger flow hierarchical control measures is as follows: the longer the information transfer, article preparation, personnel on duty, etc., the greater the N value.
S103, extracting passenger flow parameters of the service facilities and the channel facilities on the flow line every C minutes by utilizing a video identification technology;
The passenger flow parameters extracted by the service facility comprise the passenger flow arrival rate lambda i of the section upstream of the service facility, the actual queuing queue length l i of pedestrians in front of the service facility and the actual queue number r i; the passenger flow parameters extracted by the channel facilities are pedestrian walking speeds v g of 1 or more video monitoring points on each channel facility.
For a service facility, video monitoring points are distributed in a queuing space allowed area, and the saturation flow rate mu i of the ith service facility is calibrated in advance, wherein the units are as follows: people/min, namely the maximum number of people capable of passing through per unit time, the maximum allowed queuing length L i and the queuing number R i; the passenger flow arrival rate lambda i of the ith service facility is acquired in real time through a video identification technology, and the unit is as follows: people/min, i.e. the number of people actually arriving per unit time, the actual queuing captain of pedestrians l i and the actual queuing number r i, where i=1, 2,3.
For the channel facilities, P video monitoring points are distributed in key sections such as an inlet and an outlet of the channel, a turning area, a capacity bottleneck area and the like, and the pedestrian running speed of all the monitoring points is v g, the unit is: m/min, wherein: g=1, 2,3 … P, P is not less than n.
S104, calculating a passenger flow streamline congestion state evaluation index value according to the passenger flow parameters extracted by each video monitoring point, wherein the passenger flow streamline congestion state evaluation index value comprises the total average delay time T of the service facility, the occupancy rate eta of the queuing space of the whole streamline and the pedestrian travelling speed variation coefficient C v of the channel facility on the flow line;
the calculation process of the passenger flow streamline congestion state evaluation index value is as follows:
(1) Determining the distribution of the arrival rate lambda i of passengers in the peak period through survey data fitting, wherein the maximum departure rate is the saturation flow rate mu i of the facility, queuing delay is generated when lambda i is larger than mu i, and the area enclosed by the accumulated arrival and departure curves is the total queuing delay time T i of the passengers at the ith service facility, and the unit is: min, as shown in fig. 2, taking the poisson distribution obeying by the passenger arrival rate λ i as an example, the calculation formula of T i is:
The average queuing delay time t i:
Wherein: q 1 is the accumulated number of passengers arriving, and lambda i is obtained by video recognition technology according to the calculation of the distribution function of the passenger arrival rate lambda i; q 2 is the cumulative number of passengers leaving, which is the product of the saturation flow rate μ i of the ith service facility and the duration t; t is queuing duration, unit: min;
Total average delay time T of various service facilities on the streamline, unit: min, the calculation formula is:
(2) Aiming at service facilities, the full-streamline queuing space occupancy eta calculation formula is as follows:
Wherein: gamma i is the occupancy weight of queuing space of different facilities, which is obtained by expert scoring, The product of the maximum allowed queuing length L i and the number of queues R i is the maximum queuing space area allowed by station operation safety management; the pedestrian actual queuing length l i and the actual queuing number r i are obtained according to a video recognition technology;
(3) Aiming at channel facilities, the congestion degree of the channel facilities on the streamline is represented by the acquired speed fluctuation of pedestrians acquired at p channel video monitoring points, and the calculation formula of the channel pedestrian running speed variation coefficient C v is as follows:
wherein: and (5) averaging the pedestrian walking speeds of p video monitoring points of the channel facility.
S105, constructing a probabilistic neural network model PNN, taking the passenger flow streamline congestion state evaluation index value as the input of the model, evaluating the streamline congestion state in real time, and outputting N streamline congestion state evaluation grade values Y j in the streamline congestion release interval;
The structure of the probabilistic neural network is shown in fig. 3, the input vector is the total average delay time T of the service facility, the occupancy rate eta of the queuing space of the full streamline and the variation coefficient C v of the pedestrian running speed of the channel, and the neuron number of the input layer is 3; the number of neurons of the mode layer is the number of training samples; dividing the station streamline congestion level into M levels, and setting the number of neurons of the summation layer as M; the neuron of the output layer is set to 1, and the output result is the congestion level Y j with the highest probability.
The training process is as follows:
(1) Constructing a training sample data set, wherein the sample value is represented by X= (C v, T, eta), the sample size is Z, the sample size is transferred to a mode layer, and the neuron number is Z;
(2) Carrying out dimensionless processing on the passenger flow streamline congestion state evaluation index values in the sample data set;
(3) Initializing a smoothing factor, and adopting the same smoothing factor for samples belonging to the same class hidden layer;
(4) Training the probabilistic neural network model, endowing the smoothing factor to the probabilistic neural network model to start training, and calculating mode layer output, output layer output and error rate;
(5) Correcting the smoothing factor and the training threshold;
(6) Judging the error rate and the given threshold value, outputting a congestion level if the error rate is smaller than the given threshold value, entering the training of the next sample if the error exceeds 1.03 times of the previous error, otherwise, adjusting a smoothing factor;
(7) And continuously training the network until the condition is met, inputting a sample to be tested, and obtaining the streamline congestion level of the sample to be tested, wherein the streamline congestion level is divided into M levels, namely, the congestion state evaluation level value Y j = 1,2,3. j=1, 2,3.
And S106, based on N streamline congestion state evaluation grade values Y j in the streamline congestion release interval, carrying out weighted summation on Y j by taking the occurrence frequency of each streamline congestion state evaluation grade value as a weight, obtaining a streamline comprehensive congestion grade value P in the streamline congestion release interval, and releasing.
The calculation formula of the streamline comprehensive congestion level value P is as follows:
wherein: y j = 1,2, 3..m, j = 1,2, 3..n, Refers to the frequency with which each Y j occurs within the streamline congestion issue interval.
In the following, this embodiment will be described by taking a subway station in the ocean as an example, and a plan view of the station is shown in fig. 4, where the average daily passenger flow of the station is about 10 ten thousand people, the passenger flow of the double holidays is about 6 ten thousand people, and the train departure interval in the peak period is 3 min/trip. Peak passenger flow times are 8:00-9:30 and 17:00-19:00; of all the entrances, the entrance 5 has the highest passenger flow, which is about 40% of the total passenger flow, so the entrance streamline (the thickened streamline in fig. 4) taking the entrance 5 is selected as the study object, and the following steps are:
Collecting passenger flow parameters of service facilities and channel facilities;
In the selected inbound streamline, the service facility comprises: 1 entrance, 1 security inspection instrument, 1 group of entrance gate and one group of stairs; the passage facility is a pedestrian passage connected with four service facilities, and three road sections are totally provided. The passenger flow data acquisition is mainly carried out by using cameras arranged in the station, passenger flow data of each point are acquired based on videos according to a 3-min time interval by taking a time period of 6:00-23:00 as an example, 340 sample data are obtained in total, and the evaluation index values of the partial inbound streamline congestion state are shown in the following table.
TABLE 1
Outputting a congestion level;
The first 60 data are used for training the probabilistic neural network, and after the network is stable, the rest data are input into the model, and the streamline congestion level with 3 minutes as evaluation interval is obtained through the probabilistic neural network, and is shown in the table below.
TABLE 2 ingress streamline Congestion level (4 Congestion levels)
Calculating the comprehensive congestion level;
Based on the above obtained level of congestion of the flow line, the congestion state is comprehensively evaluated at 15 minutes as a distribution interval (i.e., one distribution interval includes 5 evaluation intervals), and the total congestion state is evaluated at 9:00-9:15, illustrate the calculation process of the streamline comprehensive congestion level:
The evaluation results of the 5 evaluation intervals of the distribution interval are shown in the table below, and the congestion level 4 appears 1 time and the congestion level 3 appears 3 times in the distribution interval, so that the comprehensive streamline congestion level p= [ (3×4+4×1)/5 ] =3.
9:03-9:15 Level of streamline congestion
The streamline integrated congestion level of the inbound streamline is shown in fig. 5. According to the comprehensive congestion level evaluation result, the passenger flow congestion state of the inbound streamline is intensively shown to be smooth and basically smooth in the early peak period, the congestion state can occur in a few periods, but the passenger flow can be dissipated in a short time; and during the late peak period, the passenger flow state is concentrated to be basically smooth and crowded, even severely crowded, the passenger flow on the flow line is in an extremely crowded state according to the evaluation result of the flow line congestion level at about 18:30, the station staff should immediately take measures, firstly, the source passenger flow on the flow line can be limited, namely, intermittent inbound measures are taken at the port No. 4 and the port No. 5 of the station, the arrival of passengers is reduced, secondly, the passenger flow can be guided at the outlet at the inner side of the bidirectional channel, the passenger flow at the point can be guided to the security check at the port No. 1 and the port No. 2 for ticket checking and inbound, and the passenger flow congestion caused by the passenger flow accumulation is prevented.
While particular embodiments of the present invention have been described above, it will be understood by those skilled in the art that these are by way of example only and that various changes or modifications may be made to these embodiments without departing from the spirit and scope of the invention, which is therefore defined by the appended claims.
Claims (4)
1. A real-time evaluation method for the flow line congestion state of a subway station based on video data is characterized by comprising the following steps:
(1) Selecting a main passenger flow line of a subway station in a peak period as an evaluation flow line, and dividing the constituent facilities on the flow line into m service facilities and n channel facilities according to the traffic flow characteristics of pedestrians;
(2) Setting a streamline congestion evaluation interval as C minutes and a streamline congestion release interval as NC minutes;
(3) Extracting passenger flow parameters of the service facilities and the channel facilities on the flow line every C minutes by utilizing a video identification technology;
The passenger flow parameters extracted by the service facility comprise the passenger flow arrival rate lambda i of the section upstream of the service facility, the actual queuing captain l i and the actual queuing number r i of pedestrians in front of the service facility; the passenger flow parameters extracted by the channel facilities are pedestrian running speeds v g of 1 or more video monitoring points distributed on each channel facility;
For the service facilities, arranging video monitoring points in an allowable queuing space area, and calibrating the saturation flow rate mu i of the ith service facility in advance, namely the maximum number of people passing through per unit time, the maximum allowable queuing length L i and the queuing number R i; the passenger flow arrival rate lambda i of the ith service facility, namely the number of people actually arriving in unit time, the pedestrian actual queuing captain l i and the actual queuing number r i are acquired in real time through a video identification technology, wherein i=1, 2, 3..m;
For the channel facility, P video monitoring points are distributed in key sections such as an inlet and an outlet of the channel, a turning area, a capacity bottleneck area and the like, and the pedestrian running speed of all the monitoring points is v g, wherein: g=1, 2,3 … P, P is not less than n;
(4) Calculating a passenger flow line congestion state evaluation index value according to the passenger flow parameters extracted by each video monitoring point, wherein the passenger flow line congestion state evaluation index value comprises the total average delay time T of the service facility, the occupancy rate eta of the queuing space of the whole flow line and the pedestrian travelling speed variation coefficient C v of the channel facility on the flow line;
(4.1) determining a distribution of peak passenger arrival rates lambda i by survey data fitting, the maximum departure rate being said saturation flow rate mu i for the facility, queuing delay being generated when lambda i is greater than mu i, the area enclosed between the cumulative arrival and departure curves being the total queuing delay time T i for passengers at the ith said service facility:
Average queuing delay time t i:
Wherein: q 1 is the number of accumulated passengers arriving, and lambda i is obtained by video recognition technology according to the calculation of the distribution function of the passenger arrival rate lambda i; q 2 is the cumulative number of outgoing passengers, which is the product of the saturation flow rate μ i and the duration τ of the ith service; τ is the queuing duration; the total average delay time T of various service facilities on the streamline is as follows:
(4.2) for a service facility, the full-flow line queuing space occupancy η calculation formula is:
Wherein: gamma i is the occupancy weight of queuing space of different facilities, which is obtained by expert scoring, The product of the maximum allowed queuing length L i and the number of queues R i is the maximum queuing space area allowed by station operation safety management; the actual queuing length l i and the actual queuing number r i of the pedestrians are obtained according to a video recognition technology;
(4.3) aiming at the channel facilities, representing the crowding degree of the channel facilities on the streamline by using the obtained speed fluctuation acquired by pedestrians at p channel video monitoring points, wherein the calculation formula of the channel pedestrian running speed variation coefficient C v is as follows:
wherein: the average value of the pedestrian walking speeds of p video monitoring points of the channel facility is obtained;
(5) Constructing a probabilistic neural network model PNN, taking the passenger flow streamline congestion state evaluation index value as the input of the model, evaluating the streamline congestion state in real time, and outputting N streamline congestion state evaluation grade values Y j in the streamline congestion release interval;
(6) Based on N streamline congestion state evaluation grade values Y j in the streamline congestion release interval, weighting and summing Y j by taking the occurrence frequency of each streamline congestion state evaluation grade value as a weight to obtain a streamline comprehensive congestion grade value P in the streamline congestion release interval, and releasing;
The calculation formula of the streamline comprehensive congestion level value P is as follows:
wherein: y j = 1,2, 3..m, j = 1,2, 3..n, Refers to the frequency with which each Y j occurs within the streamline congestion issue interval.
2. The real-time evaluation method for the congestion state of the flow line of the subway station based on the video data according to claim 1, wherein the method comprises the following steps: the primary passenger flow line is an inbound, outbound or transfer line with large passenger flow and 1 or more frequent congestion points.
3. The real-time evaluation method for the congestion state of the flow line of the subway station based on the video data according to claim 1, wherein the method comprises the following steps:
The service facility comprises an entrance, a security inspection machine, a gate and an escalator entrance;
The channel facilities comprise a long channel for entering and exiting the station and a long channel for transferring.
4. The real-time evaluation method of the flow line congestion state of the subway station based on the video data according to claim 1, wherein the probabilistic neural network model training process in the step (5) is as follows:
(5.1) constructing a training sample data set, wherein the sample value is represented by x= (C v, T, η), the sample size is Z, and the sample size is transferred to the pattern layer, and the neuron number is Z;
(5.2) performing dimensionless processing on the passenger flow line congestion state evaluation index value in the sample data set;
(5.3) initializing a smoothing factor, and adopting the same smoothing factor for samples belonging to the same category hidden layer;
(5.4) training the probabilistic neural network model, endowing the probabilistic neural network model with the smoothing factor to start training, and calculating a mode layer output, an output layer output and an error rate;
(5.5) modifying the smoothing factor and the training threshold;
(5.6) judging the error rate and the given threshold value, outputting a congestion level if the error rate is smaller than the given threshold value, entering training of the next sample if the error exceeds 1.03 times of the previous error, otherwise, adjusting a smoothing factor;
(5.7) continuing training the network until the condition is met, inputting a sample to be tested, and obtaining a streamline congestion level of the sample to be tested, wherein the streamline congestion level is divided into M levels, namely, the congestion state evaluation level value Y j = 1,2,3. j=1, 2,3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211620815.0A CN116229305B (en) | 2022-12-16 | 2022-12-16 | Real-time evaluation method for subway station streamline congestion state based on video identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211620815.0A CN116229305B (en) | 2022-12-16 | 2022-12-16 | Real-time evaluation method for subway station streamline congestion state based on video identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116229305A CN116229305A (en) | 2023-06-06 |
| CN116229305B true CN116229305B (en) | 2024-07-16 |
Family
ID=86590003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211620815.0A Active CN116229305B (en) | 2022-12-16 | 2022-12-16 | Real-time evaluation method for subway station streamline congestion state based on video identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116229305B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102631391B1 (en) * | 2023-07-25 | 2024-01-31 | 주식회사 공감아이티 | Crowd monitoring apparatus and method for monitoring crowd |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105321347A (en) * | 2015-09-18 | 2016-02-10 | 西安电子科技大学 | Hierarchical road network traffic jam evaluation method |
| CN107358357A (en) * | 2017-07-12 | 2017-11-17 | 北京市轨道交通设计研究院有限公司 | Urban track traffic transfer station evaluation method |
| CN110796564A (en) * | 2019-09-09 | 2020-02-14 | 国网浙江省电力有限公司经济技术研究院 | Project statistical index screening method for power grid enterprise comprehensive plan distribution measurement and calculation |
| CN110942411B (en) * | 2019-11-20 | 2022-09-09 | 北京城建设计发展集团股份有限公司 | Regional linkage based subway station passenger flow graded alarm method |
| CN115083146B (en) * | 2022-04-27 | 2023-08-29 | 同济大学 | Method for acquiring safety cause influence range in intelligent networking environment |
-
2022
- 2022-12-16 CN CN202211620815.0A patent/CN116229305B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| 基于概率神经网络的轨道交通车站客流安全状态评价;李俊辉等;《都市快轨交通》;第28卷(第4期);第1-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116229305A (en) | 2023-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110942411B (en) | Regional linkage based subway station passenger flow graded alarm method | |
| EP4030365A1 (en) | Multi-mode multi-service rail transit analog simulation method and system | |
| JP3414843B2 (en) | Transportation control device | |
| CN112598182A (en) | Intelligent scheduling method and system for rail transit | |
| CN110084397B (en) | Subway through line planning method | |
| CN108289203B (en) | Video monitoring system for rail transit | |
| CN106781499B (en) | Traffic network efficiency evaluation system | |
| CN107194491A (en) | A kind of dynamic dispatching method based on Forecasting of Travel Time between bus passenger flow and station | |
| CN104200286A (en) | Urban railway transit timetable optimizing technology application framework | |
| Xinghao et al. | Predicting bus real-time travel time basing on both GPS and RFID data | |
| CN105488751A (en) | Subway station passenger flow micro statistics and scheduling method | |
| CN108665178B (en) | AFC-based passenger flow prediction method for stairs in subway station | |
| CN114037158A (en) | Passenger flow prediction method based on OD path and application method | |
| CN113850417A (en) | Passenger flow organization decision-making method based on station passenger flow prediction | |
| CN108364464B (en) | Probability model-based public transport vehicle travel time modeling method | |
| CN116229305B (en) | Real-time evaluation method for subway station streamline congestion state based on video identification | |
| CN111401643B (en) | Urban rail transit passenger flow loop self-adaptive intelligent train scheduling method | |
| CN114707709A (en) | Safety early warning method and system for comprehensive passenger transport hub of railway | |
| CN107808235B (en) | AFC big data-based urban rail large passenger flow model construction method | |
| CN111581538A (en) | Highway charging data-based high-speed traffic flow state inference method | |
| Liu et al. | Understanding urban bus travel time: Statistical analysis and a deep learning prediction | |
| Zhou et al. | Risk analysis of crowd gathering on metro platforms during large passenger flow | |
| CN106379379A (en) | Method for predicting time spent by passengers boarding or alighting from urban rail transit | |
| CN112528867B (en) | Early warning method for sudden large passenger flow of subway station and application thereof | |
| CN115563761A (en) | Subway junction station surrounding road congestion prediction method based on timetable |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |