An intelligent passenger flow analysis method based on subway station video monitoring

Abstract

The application discloses a kind of based on metro station video monitoring crowd intelligence analysis method, the method sets decomposition frequency to carry out continuous decomposition and screening to station video data, obtains screening frame element library, sets linear activation kernel and carries out confluence extraction, obtains covering characteristic value, according to covering characteristic value, element replacement is carried out to linear activation kernel, obtains replacement capture kernel;Pass through replacement capture kernel to screening frame element library is marked capture and nuclide matching, obtains net value matching kernel, according to net value matching kernel, element net value library is constructed;To element net value library is carried out identity analogy, obtains the same kind identity net value library, net value matching kernel is carried out pattern positioning and is carried out the same kind identification with the same kind identity net value library, obtains target class identification node, to target identification node is carried out location tracking and different kind analysis, obtains traffic regulation strategy;Realize the automatic detection and tracking of crowd, improve the passenger flow management efficiency of metro station.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent video monitoring, and in particular to a passenger flow intelligent analysis method based on subway station video monitoring. BACKGROUND

[0002] As a large-capacity and high-efficiency urban traffic tool, the subway plays an important role in many large cities. However, the high passenger flow of the subway station also brings many management problems.

[0003] In order to better understand the change rule of passenger flow, reasonably arrange the vehicle shift, shorten or lengthen the departure interval, it is necessary to monitor and analyze the passenger flow in real time, so as to improve the operation efficiency and passenger satisfaction; in peak period, the subway station may appear crowded phenomenon, intelligent analysis can help the operation party to take measures in time, guide the passenger flow, prevent the occurrence of stampede; understanding the passenger flow distribution of the subway station is helpful for the subway operation party to reasonably allocate human resources and equipment resources.

[0004] Therefore, the designer of the present application, in view of the above defects, through earnest research and design, and the experience and results of many years engaged in the related industry, designs a passenger flow intelligent analysis method based on subway station video monitoring to overcome the above defects. SUMMARY

[0005] The purpose of the present application is to provide a passenger flow intelligent analysis method based on subway station video monitoring, which can effectively overcome the defects of the prior art, improve the operation efficiency of the subway station, ensure passenger safety, and reduce operation cost.

[0006] In order to achieve the above purpose, the present application discloses a passenger flow intelligent analysis method based on subway station video monitoring, characterized in that it comprises the following steps:

[0007] Step S1: constructing a station layout map and collecting station video data;

[0008] Step S2: setting a decomposition frequency to continuously decompose the station video data, obtaining continuous video frames, constructing a continuous frame element library, performing initial screening on the continuous frame element library, obtaining a screened frame element library, setting a linear activation kernel and uploading it to a marked frame element, obtaining a core coverage area, merging and extracting the core coverage area, obtaining a coverage characteristic value, replacing elements of the linear activation kernel according to the coverage characteristic value, and obtaining a replacement capture kernel;

[0009] Step S3: capturing the screened frame element library through the replacement capture kernel, obtaining a characteristic replacement element, sorting the characteristic replacement element and matching the replacement capture kernel, obtaining a net value matching kernel, and constructing an element net value library according to the net value matching kernel;

[0010] Step S4: pattern positioning is performed on the net value matching core to obtain a target mapping node, identity analogy is performed on the element net value library to obtain a same identity net value library, same identification is performed on the target mapping node through the same identity net value library to obtain a target class identification node, location tracking and different class analysis are performed on the target identification node to obtain a passing adjustment strategy.

[0011] The process of collecting the station video data includes:

[0012] The station layout diagram is constructed based on the subway station.

[0013] The subway station is monitored and deployed to obtain a video collection end.

[0014] Temporal collection is performed through the video collection end to obtain the station video data, and the collected station video data is time marked to obtain a synchronous time point.

[0015] The process of obtaining the screened frame element library includes:

[0016] The decomposition frequency is set, and the station video data is continuously decomposed according to the decomposition frequency to obtain continuous video frames.

[0017] The continuous video frames are time associated according to the decomposition frequency based on the synchronous time point to obtain frame element nodes.

[0018] The continuous frame element library is constructed based on the frame element nodes according to the continuous video frames, the obtained continuous video frames are uploaded to the continuous frame element library, the continuous frame element library is initially screened to obtain a screened frame element library, and the screened frame element library includes a marked frame element.

[0019] The process of obtaining the replacement capture core includes:

[0020] Pixel recognition is performed on the obtained screened element library to obtain frame image pixels.

[0021] The linear activation core is set according to the marked frame element, the center pixel point is obtained by center selection of the marked frame element, the linear activation core is uploaded to the marked frame element based on the center pixel point, and the corresponding area of the linear activation core in the marked frame element is marked as a core coverage area.

[0022] The core coverage area is merged and extracted to obtain a coverage characteristic value, and the linear activation core is replaced according to the coverage characteristic value to obtain a replacement capture core.

[0023] The process of merging and extracting the core coverage area includes:

[0024] The frame image pixels are expanded in the area based on the core coverage area to obtain a point-in coverage area, and the point-in coverage area is extracted in rows and columns to obtain a partial derivative coverage area.

[0025] Characteristics of the obtained partial derivative coverage are extracted to obtain coverage characteristic values.

[0026] The process of obtaining the characteristic replacement element by replacing the capture nucleus to capture the screening frame element library includes:

[0027] The capture origin is set according to the marked frame element, the replacement capture nucleus is uploaded to the capture origin of the marked frame element based on the screening frame element library, and the corresponding area of the replacement capture nucleus in the marked frame element is marked as a replacement contrast area;

[0028] The replacement contrast area is segmented by the replacement capture nucleus to obtain an extracted contrast area;

[0029] The migration step is set according to the replacement capture nucleus, the replacement capture nucleus is moved according to the migration step based on the marked frame element, and the next replacement contrast area is reached. The process of obtaining the extracted contrast area is repeated until all areas of the marked frame element are covered. The extracted contrast areas are combined according to the order of the traversal movement to obtain the characteristic replacement element.

[0030] The process of constructing the element net value library according to the net value matching nucleus includes:

[0031] The characteristic replacement element is point-ordered to obtain a net value sequence;

[0032] The net value matching nucleus is obtained by matching the net value sequence according to the obtained replacement capture nucleus;

[0033] The element net value library is constructed according to the net value matching nucleus, and the obtained net value matching nucleus is uploaded to the element net value library.

[0034] The process of obtaining the same identity net value library includes:

[0035] The video acquisition end corresponding to the net value matching nucleus is obtained, the net value position node is obtained by temporally positioning the net value matching nucleus according to the obtained video acquisition end;

[0036] The net value matching nucleus is mapped to the station layout diagram according to the net value position node to obtain a target mapping node;

[0037] The element net value library is compared according to the station layout diagram to obtain the same identity net value library.

[0038] The process of comparing the identity of the element net value library includes:

[0039] Any net value matching nucleus in the element net value library is selected as a contrast matching nucleus, and a determination result nucleus is obtained by differentiating and comparing the obtained contrast matching nucleus with the net value matching nucleus;

[0040] The same type identity net value library is obtained by checking the control matching nucleus according to the judgment result, and supplementing the same type according to the matching nucleus.

[0041] The element net value library is screened based on the same type identity net value library, and a filtered net value library is obtained.

[0042] The process of obtaining the traffic regulation strategy includes:

[0043] The target class identification node is obtained by identifying the target mapping node according to the same type identity net value library based on the station layout diagram.

[0044] The observation period is set according to the synchronization time point, the position tracking of the target class identification node is performed based on the observation period, the target cycle track is obtained, the target cycle track is analyzed, and the traffic regulation strategy is obtained.

[0045] From the above, the intelligent passenger flow analysis method based on subway station video monitoring has the following effects:

[0046] 1. Video monitoring and collection are performed on the subway station, video data is decomposed into frame images, image characters are identified, and feature information of the frame image is extracted as an identity discrimination point for track tracking, which greatly improves the data processing efficiency and the tracking speed of the target character.

[0047] 2. The station layout diagram with the same scale as the subway station is constructed to display the real-time position of the pedestrian and obtain the cycle track of the target character through the identity discrimination point, which is beneficial to the statistics of the passenger flow of the popular site, and the corresponding traffic regulation strategy can be quickly generated when the passenger flow of the popular site is too large, effectively relieving the passenger flow pressure.

[0048] The detailed content of the application can be obtained through the following description and the attached drawings. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 A schematic diagram of the intelligent passenger flow analysis method based on subway station video monitoring is shown. DETAILED DESCRIPTION

[0050] The technical solutions of the application will be described clearly and completely in combination with the embodiments. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0051] AsFigure 1 As shown, a subway station video monitoring-based passenger flow intelligent analysis method of the application is shown, which comprises the following steps:

[0052] Step S1: constructing a station layout map and collecting station video data;

[0053] Step S2: setting a decomposition frequency to continuously decompose the station video data, obtaining continuous video frames, constructing a continuous frame element library, performing initial screening on the continuous frame element library to obtain a screened frame element library, setting a linear activation kernel and uploading it to a marked frame element, obtaining a core coverage area, performing merging extraction on the core coverage area to obtain a coverage characteristic value, replacing elements of the linear activation kernel according to the coverage characteristic value to obtain a replacement capture kernel;

[0054] Step S3: performing sign capture on the screened frame element library through the replacement capture kernel to obtain a characteristic replacement element, performing point position sorting on the characteristic replacement element and performing isotope matching with the replacement capture kernel to obtain a net value matching kernel, and constructing an element net value library according to the net value matching kernel;

[0055] Step S4: performing pattern positioning on the net value matching kernel to obtain a target mapping node, performing identity analogy on the element net value library to obtain a same-class identity net value library, performing same-class identification on the target mapping node through the same-class identity net value library to obtain a target class identification node, and performing location tracking and different-class analysis on the target identification node to obtain a passing regulation strategy.

[0056] It should be further explained that, in the specific implementation process, the process of collecting station video data in step S1 comprises:

[0057] A station layout map is constructed based on a subway station, the station layout map is a three-dimensional plan view generated according to the actual structure of the subway station, and the positions of the devices and facilities of the station layout map are completely the same as those of the subway station;

[0058] The subway station is monitored and defended to obtain a video collection end, the station defense point indicates the acquisition of the overall structure of the subway station, the station defense point is set at the position that needs to be monitored, and is used to set a monitoring device to collect information, i.e., the video collection end;

[0059] The obtained video collection end is mapped to the station layout map to obtain a station defense point, wherein the station defense point is the corresponding position of the video collection end represented in the station layout map;

[0060] A transmission link between the video collection end and the station defense point is constructed, the transmission link is used to transmit the information collected by the video collection end to the station defense point;

[0061] The time point is obtained by time acquisition through a video acquisition end, and the time point is marked on the acquired station video data;

[0062] Further, the time point is marked on the start point of the acquired station video data, and the end point is also marked, and the video duration of the station video data is also marked, that is, the corresponding time point can be obtained at any time of the station video data.

[0063] In step S2, the decomposition frequency represents the frequency of decomposing the station video data, and the decomposition represents decomposing the station video data into a plurality of continuous image forms;

[0064] The station video data is continuously decomposed according to the obtained decomposition frequency, and continuous video frames are obtained;

[0065] Further, the continuous decomposition represents that each frame of the station video data is segmented according to the obtained decomposition frequency, that is, the obtained station video data is uploaded to the video processing library software, the station video data file is read through the video processing library software, the corresponding segmented frame is found according to the required frequency distribution, and the corresponding segmented frame is recorded as a continuous video frame. In particular, the video data is composed of continuous frames, that is, the continuous video frames obtained by segmenting and extracting the frames according to the decomposition frequency are in the form of images;

[0066] The continuous video frames are time-correlated according to the obtained decomposition frequency based on the synchronization time point, and frame node is obtained, wherein the time correlation represents that the station video data has a corresponding synchronization time point, and each frame in the station video data is composed in the time sequence of the synchronization time point, that is, the corresponding frame node of each continuous video frame can be obtained;

[0067] Based on the frame node, a continuous frame element library is constructed according to the obtained continuous video frames, and the obtained continuous video frames are uploaded to the continuous frame element library in the order of the frame node, that is, the arrangement order of the continuous video frames in the continuous frame element library is arranged in the time sequence;

[0068] The obtained continuous frame element library is initially screened to obtain a screening frame element library, and the screening frame element library includes a plurality of marked frame elements;

[0069] Further, the process of the initial screening includes:

[0070] The continuous frame element library is marked with a portrait to obtain a marked frame element library. The portrait marking means marking pedestrians appearing in the continuous video frames in the continuous frame element library by a pedestrian detection algorithm, that is, marking the continuous video frames in which pedestrians appear as marked frame elements and marking the continuous video frames in which no pedestrians are detected as invalid frame elements, until all the continuous video frames in the continuous frame element library are marked, to obtain the marked frame element library. The pedestrian detection algorithm includes but is not limited to a HOG feature + SVM algorithm, a DPM algorithm, a YOLO algorithm, a SSD algorithm, and a Faster R-CNN algorithm.

[0071] The marked frame element library is screened according to the obtained invalid frame elements, that is, the invalid frame elements in the marked frame element library are removed, and the marked frame elements are retained, to obtain a screened frame element library.

[0072] The screened element library is subjected to pixel recognition to obtain frame image pixels.

[0073] The pixel recognition means marking the pixel point positions in the marked frame elements in the screened element library to obtain frame image pixels, recording the obtained frame image pixels as (x, y), and recording the pixel value at the frame image pixels as a pixel normalized value.

[0074] A linear activation kernel is set according to the obtained marked frame elements. The linear activation kernel has a matrix form, the number of rows is equal to the number of columns, and is recorded as an i-row j-column matrix, that is, i = j. The matrix is composed of elements.

[0075] In particular, the elements in the set linear activation kernel can correspond to the frame image pixels in the marked frame elements at corresponding positions, that is, the linear activation kernel can adapt to the arrangement of the frame image pixels in the marked frame elements.

[0076] A center pixel point is obtained by center selection of the obtained marked frame elements. The center pixel point represents a pixel point at a center point in the marked frame elements.

[0077] The obtained linear activation kernel is uploaded to the marked frame elements based on the center pixel point, and the linear activation kernel is recorded as a core coverage area in the corresponding area of the marked frame elements, that is, the element in the i-th row and the j-th column of the linear activation kernel is coincided with the center pixel point, that is, the uploading is successful, and the frame image pixels in the core coverage area correspond to the elements in the linear activation kernel one by one. The core coverage area has the same form as the linear activation kernel, that is, the number of frame image pixels in the core coverage area is the same as the arrangement of the elements in the linear activation kernel, that is, an i-row j-column frame image pixel area.

[0078] The core coverage area is subjected to merging extraction to obtain a coverage feature value, and the linear activation kernel is replaced with the coverage feature value to obtain a replacement capture kernel.

[0079] It needs to be further explained that in the specific implementation process, the process of the merging extraction includes:

[0080] The frame image pixels in the core coverage area are marked as , (i, j) represents the i-th row and j-th column frame image pixel in the core coverage area;

[0081] The first row and first column frame image pixel in the core coverage area is expanded in the area to obtain the in-point coverage area;

[0082] The area expansion means that the surrounding elements are covered with the first row and first column frame image pixel in the core coverage area as the center to form an i-row and j-column matrix area with the first row and first column frame image pixel as the center, which is marked as the in-point coverage area, that is, the in-point coverage area is also composed of i-row and j-column frame image pixels, and the first row and first column frame image pixel in the core coverage area is the center of the in-point coverage area and the center is the i-th row and j-th column element of the in-point coverage area;

[0083] The obtained in-point coverage area is row and column extracted to obtain the partial derivative coverage area;

[0084] The process of the row and column extraction includes:

[0085] The first-order partial derivative of the frame image pixels in the in-point coverage area in the horizontal direction is calculated, that is, the first-order partial derivative in the x direction is calculated, and the obtained matrix is marked as the partial row coverage area;

[0086] The first-order partial derivative of the frame image pixels in the in-point coverage area in the vertical direction is calculated, that is, the first-order partial derivative in the y direction is calculated, and the obtained matrix is marked as the partial column coverage area;

[0087] The second-order partial derivative of the partial column coverage area is calculated according to the obtained partial row coverage area, that is, the y direction partial derivative of the partial row coverage area and the x direction partial derivative of the partial column coverage area are calculated, that is, the second-order partial derivative, and is marked as the partial derivative coverage area;

[0088] The obtained partial derivative coverage area is characterized to obtain the coverage eigenvalue, wherein the characteristic extraction means that the determinant operation is performed on the obtained partial derivative coverage area, that is, the value of the determinant of the partial derivative coverage area is calculated, and the obtained value of the determinant is marked as the coverage eigenvalue;

[0089] By analogy, the first row and second column frame image pixel in the core coverage area is expanded in the area, and the process of repeatedly obtaining the coverage eigenvalue is repeated, and the first row and third column frame image pixel in the core coverage area is expanded in the area, the first row and fourth column frame image pixel in the core coverage area is expanded in the area, and so on, until all frame image pixels in the core coverage area and the corresponding coverage eigenvalue are obtained;

[0090] According to the obtained coverage characteristic value, the obtained linear activation kernel is element replaced to obtain a replacement capture kernel;

[0091] Further, the element replacement process includes:

[0092] According to the coverage characteristic value being obtained corresponding to the frame image pixel point of the first row and the first column, the frame image pixel point of the first row and the second column, the frame image pixel point of the first row and the third column, the frame image pixel point of the first row and the fourth column, and so on in the core coverage area, the element in the first row and the first column in the linear activation kernel is replaced into the coverage characteristic value corresponding to the frame image pixel point of the first row and the first column in the core coverage area, the element in the first row and the second column in the linear activation kernel is replaced into the coverage characteristic value corresponding to the frame image pixel point of the first row and the second column in the core coverage area, and so on until all the elements in the linear activation kernel are replaced to obtain the replacement capture kernel.

[0093] The obtained replacement capture kernel can be uploaded to the screening frame element library, and the screening frame element library is captured by the replacement capture kernel to obtain a characteristic replacement element.

[0094] It needs to be further explained that in the specific implementation process, the process of the mark capture in step S3 includes:

[0095] According to the obtained mark frame element, a capture origin is set, and the obtained replacement capture kernel is uploaded to the capture origin of the mark frame element based on the screening frame element library, wherein the capture origin represents a point set in the mark frame element for determining the initial moving position of the replacement capture kernel, that is, the first mark frame element to which the replacement capture kernel is uploaded in the screening frame element library, and the corresponding area of the replacement capture kernel in the mark frame element is marked as a replacement contrast area.

[0096] The replacement contrast area is segmented extracted by the replacement capture kernel to obtain an extracted contrast area.

[0097] Further, the segmentation extraction means that the elements in the replacement capture kernel are convolved with the pixel normalization values of the corresponding frame image pixel points in the replacement contrast area, and the corresponding position elements are convolved to obtain the extracted contrast area, and the pixel normalization values of the frame image pixel points in the extracted contrast area are replaced by the values after the convolution.

[0098] According to the obtained replacement capture kernel, a migration step is set, the migration step represents the moving interval of the replacement capture kernel in the mark frame element, that is, the migration step, in order to improve the feature extraction accuracy of the mark frame element, the migration step needs to be small enough, that is, the migration step is equal to half of the number of rows of the replacement capture kernel, that is, each time the replacement capture kernel is moved by the interval of one frame pixel point.

[0099] According to the obtained replacement capture kernel, the obtained replacement capture kernel is traversed and moved according to the obtained migration step length, a next replacement contrast area is reached, a section is extracted from the replacement contrast area by the replacement capture kernel, and the extraction is performed until all areas of the mark frame element are covered. The obtained extracted contrast areas are combined in sequence according to the traversal and movement, a characteristic replacement element is obtained, the obtained characteristic replacement element has the same form as the mark frame element, the frame image pixel position is unchanged, only the pixel normalization value at the frame image pixel position is changed, which is the value after convolution with the replacement capture kernel, and the pixel normalization value after convolution is recorded as a replacement normalization value;

[0100] Further, the traversal and movement means that the replacement capture kernel at the capture origin position is moved by a migration step length, a next replacement contrast area is reached, and a section is extracted from the replacement contrast area, an extracted contrast area is obtained, and the sequence of the traversal and movement is from left to right, and after reaching the boundary of the mark frame element, the position is moved by one height of the replacement capture kernel and then moved to the right.

[0101] In particular, if the replacement capture kernel exceeds the boundary of the mark frame element after being moved by the migration step length, the replacement capture kernel is only moved to the position where the replacement capture kernel coincides with the boundary of the mark frame element.

[0102] The obtained characteristic replacement element is point-position sorted to obtain a net value sequence.

[0103] The point-position sorting means that the values of the frame image pixels in the characteristic replacement element are sorted in descending order to obtain a net value sequence, and the net value sequence is arranged for the frame image pixels.

[0104] The net value sequence is matched with the obtained replacement capture kernel to obtain a net value matching kernel.

[0105] The nuclide matching means that according to the arrangement sequence of the replacement capture kernel, that is, an i-row j-column matrix, an i-row j-column matrix is constructed, which is recorded as an original sequence kernel, and the constructed matrix is also an equal-row equal-column matrix. In the net value sequence, the top i x j frame image pixels are selected, and the selected frame image pixels are sequentially uploaded to the constructed original sequence kernel, and the corresponding values of the frame image pixels are reserved at the corresponding element positions of the original sequence kernel to obtain a net value matching kernel. The obtained net value matching kernel is a matrix formed by the replacement normalization values of the top i x j frame image pixels in the net value sequence.

[0106] The obtained net value matching kernel is associated with the corresponding mark frame element.

[0107] According to the obtained net value matching kernel, an element net value library is constructed, and the obtained net value matching kernel is uploaded to the element net value library.

[0108] The video acquisition end corresponding to the net value matching check is acquired, the net value matching check is temporally positioned according to the acquired video acquisition end, and a net value position node is obtained;

[0109] In step S4, the temporal positioning represents that the station video data corresponding to the net value matching check is acquired, the video acquisition end corresponding to the station video data is obtained, the position information corresponding to the net value matching check is determined through the video acquisition end in the monitoring defense position of the subway station, and is recorded as a net value position node, and the obtained net value position node is associated with the corresponding net value matching check;

[0110] The obtained net value matching check is mapped to the station layout diagram according to the obtained net value position node, a target mapping node is obtained, and the obtained net value matching check is uploaded to the corresponding target mapping node;

[0111] The target mapping node represents that the net value position node is uploaded to the station layout diagram through a transmission link and is mapped to the same position to obtain a target mapping node, and the target mapping node has a corresponding associated net value matching check, that is, the net value matching check represents the identity feature information corresponding to the mark frame element, and all the net value position nodes corresponding to the obtained net value matching checks are mapped to the station layout diagram;

[0112] Identity analogy is performed on the obtained element net value library based on the station layout diagram to obtain a same-class identity net value library;

[0113] It needs to be further explained that, in the specific implementation process, the process of identity analogy includes:

[0114] An optional net value matching check in the element net value library is recorded as a contrast matching check, an optional net value matching check in the remaining net value matching checks is obtained, the obtained contrast matching check is differentially compared with the net value matching check, and a judgment result nucleus is obtained;

[0115] Further, the process of differential comparison includes:

[0116] The contrast matching check and the corresponding position elements in the net value matching check are subjected to difference operation, that is, the contrast matching check and the corresponding position elements in the net value matching check are subtracted to obtain an analogy nucleus;

[0117] The obtained analogy nucleus is subjected to element judgment to obtain a judgment result nucleus, and the judgment result nucleus includes a matching type nucleus and a non-conforming type nucleus;

[0118] The element judgment represents that the elements in the analogy nucleus are judged one by one, when all the elements in the analogy nucleus are zero, the analogy nucleus is recorded as a matching type nucleus, and when there are elements other than zero in the analogy nucleus, the analogy nucleus is recorded as a non-conforming type nucleus;

[0119] In the residual net value matching core, another net value matching core is selected, and differential comparison is performed between the selected net value matching core and the control matching core until differential comparison is completed between the residual net value matching core and the control matching core;

[0120] Based on the element net value library, the same type supplementary is performed on the control matching core according to the judgment result, and a same type identity net value library is obtained;

[0121] The same type supplementary means that when the judgment result core is a matching type core, the same type identity net value library is constructed according to the net value matching core corresponding to the matching type core and the control matching core, and the net value matching core corresponding to the matching type core and the control matching core are uploaded to the same type identity net value library, that is, the identities of the persons in the net value matching core corresponding to the matching type core and the control matching core are the same, that is, the identities of the pedestrians in the same type identity net value library are completely the same as the pedestrian information of the control matching core, but the positions captured and the behaviors performed are different at different times;

[0122] Based on the same type identity net value library, multi-source screening is performed on the element net value library, and a filtered net value library is obtained;

[0123] The multi-source screening means that the net value matching core included in the same type identity net value library is removed from the element net value library;

[0124] In the obtained filtered net value library, a net value matching core is selected as a control matching core, and the process of differential comparison of the control matching core is repeated until all the net value matching cores in the element net value library are subjected to differential comparison, and type distribution is completed, the type distribution means that all the net value matching cores in the element net value library can be matched to corresponding matching type cores, that is, the same type identity net value library is obtained, that is, one same type identity net value library represents the image information of one pedestrian in the time length of the subway station video data in the subway station;

[0125] Based on the station layout map, the same type identity net value library is used to perform same type identification on the target mapping node, and a target class identification node is obtained;

[0126] The same type identification means that in the station layout map, the target mapping nodes associated with the net value matching cores in the same same type identity net value library are marked with the same identification point, and the target class identification node is obtained; for example, the same color is marked, and whether the pedestrians are the same person can be distinguished by observing the color of the target mapping node in the station layout map;

[0127] According to the obtained synchronization time point, an observation period is set, which is expressed as a longer time period, according to the operation rules of the subway station and the monitoring demand, the synchronization time point is combined to set a longer time period as the observation period, and the time period should be able to cover multiple passenger flow peaks and valleys;

[0128] Based on the observation period, the position tracking of the target class identification node is performed to obtain a target periodic track;

[0129] It should be further explained that, in the specific implementation process, the position tracking process includes:

[0130] According to the time sequence of the frame element nodes corresponding to the marked frame elements of the net value matching core corresponding to the target class identification node, the target class identification node in the station layout diagram is virtually connected to obtain the movement track of the target pedestrian corresponding to the same identity net value library in the subway station, that is, the target periodic track, which represents the movement track in the subway station within a period of time, and the riding track of the target pedestrian is known through the station layout diagram, wherein the target pedestrian represents a pedestrian corresponding to a same identity net value library;

[0131] The target periodic track obtained is subjected to heterogeneous analysis to obtain a traffic regulation strategy;

[0132] The heterogeneous analysis represents an analysis of the path coincidence of the target periodic tracks corresponding to different same identity net value libraries;

[0133] Based on the observation period, the coincidence degree of the target periodic tracks corresponding to different same identity net value libraries is calculated to obtain a track coincidence degree, the track coincidence degree represents that the repeated target class identification nodes in all target periodic tracks within an observation period are counted to obtain a coincidence node number, wherein the coincidence node number represents the number of times that a same target class identification node appears in different target periodic tracks;

[0134] According to the obtained coincidence node number, a node number threshold is set, and the node number threshold is used to generate a traffic regulation strategy;

[0135] Then, according to the obtained traffic regulation strategy, the traffic flow pressure at the net value position node is relieved, the node number threshold represents a threshold number set according to the real-time running condition of the subway station, and is used to generate a traffic regulation strategy, the traffic regulation strategy represents that when the coincidence node number of a target identification node is greater than or equal to the node number threshold, a shunting strategy is generated according to the coincidence node number of other target identification nodes, the people flow at the target identification node is dispersed to other target identification nodes, and a traffic regulation strategy is obtained.

[0136] Therefore, the advantages of the present application are that:

[0137] By decomposing the collected video data into image form, marking the pixel points in the image, and setting the extraction kernel, the second-order partial derivative matrix is generated by the center pixel point to update the elements of the extraction kernel, and the replacement capture kernel for extracting features is obtained, which is convenient for extracting various details in the image, enhancing the image content, and reducing misjudgment and missed judgment;

[0138] Then the updated capture kernel is used to slide capture the pixels in the image, obtain the feature pixel value after feature extraction, and sort the feature pixel value to obtain the net value sequence. According to the net value sequence, the net value matching kernel corresponding to each image is constructed. By setting key features for targeted feature extraction, it helps to improve the reliability and accuracy of image feature extraction, so that the computer can more accurately understand the information in the image, and it is convenient for the management personnel to set specific target features to search for target passengers in the station, and increase the search speed. For example, subway managers need to find passengers with yellow hats and red backpacks. Then set the extraction kernel and update the elements to obtain the replacement capture kernel to extract the net value matching kernel related to it in the image, accurately grasp the position information and behavior trajectory of the target passenger, help the staff to make quick decisions and take effective emergency measures to protect the safety of passengers.

[0139] According to the obtained net value matching kernel, a feature image library generated by each passenger in the subway station is constructed, a filtered net value library is obtained, which contains all image information of a passenger in the subway station within the time length of a station video data, and the position node of each net value matching kernel in the filtered net value library is marked in the station layout map. According to the time sequence, all position nodes in the filtered net value library are connected to obtain the action trajectory of the passenger within an observation period. It can more accurately discover the passenger flow rules in the subway station, such as peak hours and popular areas, which helps the subway operation department to reasonably arrange staff and adjust operation strategies to improve operation efficiency. At the same time, according to the action trajectory of passengers in the subway station, when a sudden event occurs, clear images can provide accurate passenger flow information in time to help staff make quick decisions and take effective emergency measures to protect passenger safety.

[0140] Finally, the repeated position nodes in the action trajectories of all passengers within an observation period are counted to obtain the number of coincident nodes, and the size of the coincident node is used to determine the passenger flow size of the position node, and then a traffic regulation strategy is generated to provide a reference for the planning and optimization of urban rail transit network, reasonably adjust the line direction, station setting, etc. Improve the coverage range and service level of rail transit network and promote the sustainable development of urban traffic.

[0141] It is expressly understood that the description and specific examples thereof are only given by way of example and are not intended to limit the disclosure, application or use of the application. While the application has been described in connection with specific embodiments thereof, it will be understood that the application is capable of further modifications. The application is intended to cover any variations, uses or adaptations of the application including such departures from the present disclosure as come within known or customary practice in the art to which the application pertains and falls within the scope of the appended claims.

Claims

1. A people flow intelligent analysis method based on subway station video monitoring, characterized in that The method comprises the following steps: Step S1: constructing a station layout map and collecting station video data; Step S2: setting a decomposition frequency to continuously decompose the station video data, obtaining continuous video frames, constructing a continuous frame element library, performing initial screening on the continuous frame element library to obtain a screened frame element library, setting a linear activation kernel and uploading the linear activation kernel to a marked frame element to obtain a core coverage area; Based on the core coverage area, the frame image pixels are expanded in the area to obtain an in-point coverage area, and the in-point coverage area is extracted in rows and columns to obtain a partial derivative coverage area; The row and column extraction means that the first-order partial derivative of the frame image pixels in the in-point coverage area is calculated in the horizontal direction to obtain a partial row coverage area, and the first-order partial derivative of the frame image pixels in the in-point coverage area is calculated in the vertical direction to obtain a partial column coverage area, and the second-order partial derivative of the partial row coverage area with respect to the partial column coverage area is calculated to obtain the partial derivative coverage area; The obtained partial derivative coverage area is subjected to feature extraction to obtain a coverage characteristic value; The coverage characteristic value is used to replace elements of the linear activation kernel to obtain a replacement capture kernel; Step S3: setting a capture origin according to the marked frame element, uploading the replacement capture kernel to the capture origin of the marked frame element based on the screened frame element library, and marking the corresponding area of the replacement capture kernel in the marked frame element as a replacement contrast area; The replacement contrast area is subjected to section extraction by the replacement capture kernel to obtain an extracted contrast area; A migration step is set according to the replacement capture kernel, and the replacement capture kernel is moved according to the migration step based on the marked frame element to reach the next replacement contrast area, and the process of obtaining the extracted contrast area is repeated until all areas of the marked frame element are covered, the extracted contrast areas are combined according to the order of the movement to obtain a characteristic replacement element; The characteristic replacement element is subjected to point position sorting and nuclide matching with the replacement capture kernel to obtain a net value matching kernel, and the net value matching kernel is used to construct an element net value library; Step S4: positioning the net value matching kernel to obtain a target mapping node, comparing the identity of the element net value library to obtain a same-identity net value library, identifying the target mapping node through the same-identity net value library to obtain a target class identification node, tracking the position of the target identification node and performing different-class analysis to obtain a passing regulation strategy. 2.The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of collecting station video data comprises: constructing a station layout map based on a subway station; monitoring and defending the subway station to obtain a video collection end; temporally collecting the station video data through the video collection end to obtain the station video data, and time-marking the collected station video data to obtain a synchronous time point. 3.The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of obtaining the screened frame element library comprises: setting a decomposition frequency, continuously decomposing the station video data according to the decomposition frequency to obtain continuous video frames; time-correlating the continuous video frames according to the decomposition frequency based on the synchronous time point to obtain a frame element node; constructing a continuous frame element library according to the continuous video frames based on the frame element node, uploading the obtained continuous video frames to the continuous frame element library, performing initial screening on the continuous frame element library to obtain a screened frame element library, and the screened frame element library comprises a marked frame element.

4. The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of obtaining the replacement capture kernel comprises: Pixel recognition is performed on the obtained screening element library to obtain frame image pixel points; A linear activation kernel is set according to the marked frame element, a center pixel point is obtained by center selection on the marked frame element, the linear activation kernel is uploaded to the marked frame element based on the center pixel point, and the linear activation kernel is marked as a core coverage area in the corresponding area of the marked frame element; The obtained core coverage area is merged and extracted to obtain a coverage characteristic value, the linear activation kernel is replaced according to the coverage characteristic value, and a replacement capture kernel is obtained.

5. The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of constructing the element net value library according to the net value matching kernel includes: Point position sorting is performed on the characteristic replacement element to obtain a net value sequence; Nucleus matching is performed on the net value sequence according to the obtained replacement capture kernel to obtain a net value matching kernel; The element net value library is constructed according to the net value matching kernel, and the obtained net value matching kernel is uploaded to the element net value library.

6. The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of obtaining the same type identity net value library includes: A video acquisition end corresponding to the net value matching kernel is obtained, and the net value matching kernel is time-positioned according to the obtained video acquisition end to obtain a net value position node; The net value matching kernel is mapped to the station layout diagram according to the net value position node to obtain a target mapping node; Identity analogy is performed on the element net value library based on the station layout diagram to obtain the same type identity net value library.

7. The intelligent passenger flow analysis method based on subway station video monitoring according to claim 1, characterized in that, The process of performing identity analogy on the element net value library includes: A net value matching kernel in the element net value library is selected as a control matching kernel, difference comparison is performed between the obtained control matching kernel and the net value matching kernel to obtain a judgment result kernel; The same type identity net value library is obtained by performing same type supplement on the control matching kernel based on the element net value library according to the judgment result kernel; Multi-source screening is performed on the element net value library based on the same type identity net value library to obtain a filtered net value library, a net value matching kernel in the filtered net value library is selected as a control matching kernel, and the process of obtaining the same type identity net value library is repeated until all net value matching kernels in the element net value library are completed type distribution. 8.The intelligent passenger flow analysis method based on subway station video monitoring of claim 1, wherein, The process of obtaining the passing regulation strategy includes: Same type identification is performed on the target mapping node based on the station layout diagram according to the same type identity net value library to obtain a target type identification node; An observation period is set according to a synchronous time point, position tracking is performed on the target type identification node based on the observation period to obtain a target period track, and the passing regulation strategy is obtained by performing different type analysis on the target period track.

Images