Intelligent passenger flow saturation analysis method based on scenic spot multi-channel monitoring video

Abstract

The invention relates to the technical field of video big data analysis and calculation, in particular to an intelligent passenger flow saturation analysis method based on scenic spot multi-channel monitoring videos. According to the method, the effective park entering condition is judged according to double judgment bases, and finally the passenger flow saturation is calculated according to the maximum bearing capacity of the scenic spot. Firstly, whether tourists enter and leave the entrance and exit range of the scenic spot is determined through tourist entrance and exit boundaries; secondly, comparing a passenger flow statistical area angle threshold value with a straight line angle formed by intersection of a tourist traveling track and a passenger flow statistical area oblique angle value, and judging whether the tourist behavior is one-time effective entering; in addition, whether the tourists repeatedly enter the park is judged according to the face feature vectors of the tourists, and whether one-time park entering and leaving behaviors of the tourists are calculated as effective passenger flow is comprehensively judged. Compared with a traditional passenger flow statistics mode in the market, the method has the advantages that the scenic spot passenger flow can be more accurately counted through face recognition and tourist track comprehensive judgment, and then the scenic spot passenger flow saturation is calculated.

Description

Technical Field

[0001] The present invention relates to the field of video big data analysis and computing technology, and in particular to an intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of a scenic spot. Background Art

[0002] With the gradual recovery of the tourism market and the improvement of people's living standards, the tourism market has brought many challenges to scenic spots in terms of operation, service, and management. For example, crowded tourists in scenic spots cause safety hazards, long queues cause poor travel experience, etc. In order to improve the quality of tourist services and to enable scenic spots to conduct refined passenger flow management, help scenic spots quickly understand the passenger flow carrying pressure, formulate and implement passenger flow control plans in advance, it is crucial to intelligently, real-time and accurately calculate the passenger flow saturation of scenic spots. Summary of the invention

[0003] The present invention relates to an intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of a scenic spot. During the opening period of the scenic spot, multi-channel surveillance videos in the scenic spot are used to set a passenger flow statistical area, identify and calculate tourists in the passenger flow statistical area, thereby obtaining the effective passenger flow reception capacity of the scenic spot, and then according to the "Guidelines for Determining the Maximum Carrying Capacity of Scenic Spots" and the effective passenger flow reception capacity of the scenic spot, the passenger flow saturation of the scenic spot is analyzed, calculated, and output. The advantages of the present invention are: relying on the existing multi-channel surveillance videos installed in the scenic spot, the effective passenger flow of the scenic spot to be monitored is obtained through area setting, portrait comparison, and trajectory analysis, so as to calculate the passenger flow saturation.

[0004] The present invention provides an intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of scenic spots. The method calculates the number of people in the scenic spot by performing face recognition and trajectory tracking through surveillance videos at the entrances and exits of the scenic spot, and intelligently analyzes and calculates the passenger flow saturation according to the maximum tourist carrying capacity calculated by each scenic spot under the guidance of the "Guidelines for Determining the Maximum Carrying Capacity of Scenic Spots", thereby providing data support for the adaptive management of the scenic spot in terms of ticket reservation, diversion and guidance, early warning reporting, special plans, etc.

[0005] The present invention is achieved through the following technical solutions: The intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of scenic spots includes the following steps: S1. Install monitoring equipment at the entrance and exit of the scenic area or select installed monitoring equipment to record monitoring video in real time; S2. Set the passenger flow statistics area, calculate and enter the standard value of the moving trajectory: 1. Set the passenger flow statistics area. Select an area from the entire screen area of ​​the surveillance video recorded in S1 as the passenger flow statistics area. The passenger flow statistics area consists of the travel channel, entry boundary and exit boundary. The monitoring equipment records the entire process of tourists entering the entry boundary and leaving the exit boundary. 2. Calculate and enter the standard value of the moving trajectory, which is composed of the oblique angle value of the statistical area and the angle threshold of the statistical area; (1) Statistical area oblique angle value: select any point P from the entry boundary line in the direction of tourists entering, draw a straight line L1 about point P according to the horizontal angle of the video screen, and then draw a straight line L2 perpendicular to the entry boundary line. The oblique angle value of the statistical area is ∠eng. This oblique angle value represents the general direction of tourists in the entire system. (2) Angle threshold of the statistical area: Draw two diagonal lines in the designated passenger flow statistical rectangular area. The angles between the two diagonal lines and the travel channel are ∠Le and ∠Re respectively. The angle threshold of the statistical area is ∠Rp. Set ∠Rp=∠Le=∠Re. In the recorded surveillance video, P0 is the point where tourists enter the access boundary of the passenger flow statistics area; Pn is the point where tourists leave the exit boundary of the passenger flow statistics area; draw a horizontal line L1 about point P0 based on the video screen; the horizontal line L1 intersects with the vertical line L2 of the rectangular area about the access boundary, and the intersection angle is the oblique angle value ∠eng of the statistical area; connect the two points P0 and Pn and extend them to obtain the line segment L3, and record the intersection angle of L3 and L2 as ∠pen, and compare the value range of ∠pen with the angle threshold ∠Rp of the statistical area; S3, face recognition and recording of face feature vectors, and removal of duplicate passenger flows by feature vector comparison. First, face detection technology is used to obtain feature vectors corresponding to all faces in the video stream, and the Euclidean distance between faces is set. If the Euclidean distance between two faces is less than the set threshold sim, they are considered to be the same person. S4: extract moving targets and track trajectories, compare the angles of the starting and ending points of the moving trajectories with the standard values ​​of the moving trajectories, and use face detection and tracking algorithm technology to obtain the location information of the tourists entering the entry boundary of the passenger flow statistical area and the location information of the tourists leaving the exit boundary of the passenger flow statistical area. Extract moving targets and track trajectories. When tourists enter the entry boundary of the passenger flow statistical area, use face detection technology to identify the portrait of the tourists, calculate the feature vector of the face and save it. In addition, the portrait of the entry boundary is taken as the first frame, and the position of the tourists is recorded as P0. Next, the position of the tourists in each frame is obtained and recorded in the set area until the tourists leave the exit boundary of the passenger flow statistical area and record it as the final exit position Pn. The positions of the tourists in each frame are connected in sequence to obtain the direction of travel of the tourists in the passenger flow statistical area. S5: Comprehensively judge the entry and exit situation based on the feature vector and the movement trajectory, firstly perform deduplication judgment based on the saved facial feature vector of the tourist; 1. Obtain the facial feature vector of the tourist who enters the passenger flow statistics area on that day, and compare it with the facial feature vector saved on that day. If the Euclidean vector is greater than the threshold sim, it is considered that the same person does not need to be counted; if the Euclidean vector is less than the threshold sim, it is considered that the tourist did not enter the park on that day; 2. According to step S4, the intersection angle ∠pen determined by the visitor's entry and exit location information and the statistical area angle threshold ∠Rp are used for judgment, and the action trajectory where ∠pen is less than or equal to ∠Rp is taken; 3. If both of the above conditions are met, the visitor is considered to have successfully entered or left the park. Otherwise, the visitor's trajectory is considered abnormal and will not be counted in the visitor flow. S6: Count the passenger flow of the current device for effective entry and exit. According to the above steps, the passenger flow data is counted according to different camera points. For the scenic spot, the entrance and exit cameras are located in different positions. Information is provided for the position and the scenic spot, and the entrance and exit of the surveillance camera are distinguished. Finally, the calculation formula for the real-time number of people in the park E is obtained as follows: In the formula, En is the passenger flow counted by a single entrance surveillance video, and Eg is the passenger flow counted by a single exit surveillance video. The passenger flow counted by all selected entrance surveillance devices and all exit surveillance devices is calculated, and the result is the real-time number of people in the park; S7: Calculate the saturation of the scenic spot based on the passenger flow statistics and output it. According to the maximum carrying capacity determined by the scenic spot and the passenger flow entering and leaving the scenic spot counted by the multi-channel surveillance video of the scenic spot, the passenger flow saturation formula of the scenic spot can be finally determined as follows: R = E / (Max SC – k)*100 In the formula: R is the tourist flow saturation of the scenic spot; k is the scenic area staff and needs to be eliminated from the scenic area tourists; E is the real-time number of people in the scenic area; Max SC It is the maximum carrying capacity of the scenic area.

[0006] In S1, monitoring equipment is installed at the entrance and exit of the scenic area or installed monitoring equipment is selected. The specific technical requirements of the monitoring equipment are as follows: The resolution of the monitoring equipment is above 1080P; the monitoring equipment is located above the ticket gate, queuing area or entrance and exit of the scenic area, at a height of 2.5-3.5 meters, and records the monitoring video in real time.

[0007] The statistical area angle threshold ∠Rp in S2 ranges from 10 to 30°.

[0008] Beneficial effects of the present invention: The present invention relates to an intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of scenic spots. This method calculates the real-time number of people in the scenic spot through surveillance videos at the entrances and exits of the scenic spot. Compared with the common video passenger flow statistics method, this method calculates a more accurate real-time number of people in the park by defining passenger flow statistics areas, portrait feature vectors, etc., and calculates and outputs a more accurate passenger flow saturation according to the maximum carrying capacity issued by the scenic spot under the guidance of the "Guidelines for Determining the Maximum Carrying Capacity of Scenic Spots", thereby providing data basis for the scenic spot in terms of guidance, control, etc. The technical points and beneficial effects of the invention content to be protected are as follows; The present invention is about the passenger flow saturation analysis method, which adopts a double judgment basis to judge the effective entry into the park, and finally calculates the passenger flow saturation according to the maximum carrying capacity of the scenic spot. First, it determines whether the tourists enter or leave the entrance and exit range of the scenic spot through the tourist entry and exit boundaries, and then compares the angle threshold of the passenger flow statistical area and the straight line angle of the intersection of the tourist's travel trajectory and the oblique angle value of the passenger flow statistical area to judge whether the tourist behavior is a valid entry into the park; in addition, it judges whether the tourist is repeatedly entering the park according to the facial feature vector of the tourist. According to the above two judgment bases, it is comprehensively judged whether the tourist's entry and exit behavior is counted as effective passenger flow. The advantage of the present invention over the traditional passenger flow statistics method on the market is that it can more accurately count the passenger flow of the scenic spot through comprehensive judgment of face recognition and tourist trajectory, and then calculate the passenger flow saturation of the scenic spot. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Set up a schematic diagram of the passenger flow statistics area for the scenic area channel; Figure 3 A schematic diagram of the oblique angle values ​​for tourists entering and leaving the statistical area; Figure 4 A schematic diagram of the oblique angle value of the scenic spot entrance video and picture channel statistics area; Figure 5 A schematic diagram of a rectangular area for passenger flow statistics in video images; Figure 6 Information on the points where tourists enter and leave the passenger flow statistics area; Figure 7 A schematic diagram of the direction of travel of tourists in the passenger flow statistics area; Figure 8 A schematic diagram for setting entry and exit boundaries; Fig. 9 It is the overall implementation of passenger flow saturation and the schematic diagram of passenger flow saturation output. DETAILED DESCRIPTION

[0010] The technical solution of the present invention is further described with reference to the accompanying drawings: Figure 1The passenger flow calculation method for a single monitoring device is described in detail below.

[0011] S1: Among the monitoring devices installed at the entrances and exits of the scenic area, select the monitoring devices with clear images and complete coverage of the entrances and exits. The specific selection is described as follows: 1. Choose high-resolution monitoring equipment, at least 1080P (1920x1080) or above. The higher the resolution, the clearer the picture.

[0012] 2. At a height of 2.5 to 3.5 meters above the entrance and exit of the scenic area, it can cover the people entering and leaving to the maximum extent and prevent tourists from blocking each other.

[0013] 3. The monitoring equipment area ensures effective coverage of key areas such as ticket gates, queuing areas, and gate entrances and exits.

[0014] 4. Select the direction directly opposite to the tourists’ source direction to facilitate more accurate identification of tourists’ facial feature vectors for tracking tourists’ trajectories, thereby more accurately counting effective inbound and outbound passenger flow.

[0015] S2: Set the passenger flow statistics area, calculate and enter the standard value of the movement trajectory. This step is divided into two steps, which are explained in detail below.

[0016] 1. Set the passenger flow statistics area. The video image of the monitoring device selected in step S1 includes not only the entrance and exit of the scenic area, but also the squares and roads around the entrance and exit. The behavior of people in these areas has nothing to do with the flow of people entering and leaving the scenic area. In order to reduce the scope of face recognition and trajectory tracking in the monitoring video screen, a specific area is selected from all areas of the monitoring screen as the passenger flow statistics area, such as Figure 2 The locale settings are described below.

[0017] (1) The lengths of the entry and exit boundaries of the passenger flow statistics area are set according to the actual width of the scenic area entrance and exit.

[0018] (2) The distance between the entry and exit boundaries of the passenger flow statistics area is set within a time range of 4 to 8 seconds, that is, 6 to 10 meters, based on the normal walking speed of 1.4 meters per second.

[0019] (3) Taking entrance to the park as an example, the entry boundary of the passenger flow statistics area is outside the scenic area entrance, and the exit boundary of the passenger flow statistics area is inside the scenic area entrance.

[0020] like Figure 2For the exit passage of a scenic spot, a camera located above the gate of the scenic spot and tilted downward was selected based on field investigation. The rectangular area in the figure is the passenger flow statistics area. The length of the entry boundary and exit boundary line segments of the statistics area is the width of the passage. The length is set at 6 meters according to the above requirements, that is, the middle section of the entrance passage is selected.

[0021] 2. Calculate and enter the standard value of the moving trajectory. This value consists of two parts: the oblique angle value of the statistical area and the angle threshold of the statistical area. It is mainly used to determine the direction of movement of personnel after identifying their moving trajectory. The specific description and settings are as follows.

[0022] (1) Statistical area oblique angle values, as follows Figure 3 As shown, select any point P from the entry boundary line in the enter direction, make a straight line L1 about point P according to the horizontal angle of the video screen, and then make a straight line L2 perpendicular to the entry boundary line. The oblique angle value of the statistical area is ∠eng. Because the oblique angle value represents the general direction of tourists' travel in the entire system, it is clearly defined as the horizontal coordinate axis X-axis is 0 degrees from the system design and planning stage, and increases successively to the fourth quadrant, the third quadrant, the second quadrant, and the first quadrant.

[0023] according to Figure 3 Description, such as Figure 4 It can be concluded that the oblique angle value ∠eng of the entrance channel of a certain scenic spot is 30 o The angle value can be determined by using the angle measurement tool or the angle measurement tool that comes with the mobile phone compass.

[0024] (2) Statistical area angle threshold, as follows Figure 5 As shown in the figure, two diagonal lines are drawn in the designated passenger flow statistics rectangular area, and the lower left ∠Le and the lower right ∠Re are recorded respectively. By using the rectangular feature and triangle congruence judgment, it can be determined that ∠Le=∠Re, and the angle threshold of the statistical area is recorded as ∠Rp, and ∠Rp=∠Le=∠Re. The angle threshold of the departure statistical area of ​​the scenic spot is determined to be 10 o Up to 30 o .

[0025] exist Figure 6 In the figure, P0 is the point where tourists enter the entry boundary of the passenger flow statistics area. Pn is the point where tourists leave the exit boundary of the passenger flow statistics area.

[0026] (1) P0 is the point where tourists enter the passenger flow statistics area. A horizontal line L1 is drawn about P0 based on the video image.

[0027] (2) With respect to the horizontal line L1, make the horizontal line L1 intersect with the perpendicular line L2 of the rectangular area about the access boundary. The intersection angle is the oblique angle value ∠eng of the statistical area.

[0028] (3) Connect points P0 and Pn and extend them to obtain line segment L3. The angle at which L3 and L2 intersect is ∠pen. The value range of ∠pen is compared with the angle threshold ∠Rp of the statistical area. There are three cases, as shown in Table 1.

[0029] Table 1 Relationship between judgment basis and judgment result S3: This step is divided into two steps: face recognition and recording of face feature vectors, and removing duplicate passenger flows by feature vector comparison. First, face detection technology is used to obtain feature vectors corresponding to all faces in the video stream, and the Euclidean distance between faces is set. If the Euclidean distance between two faces is less than the set threshold sim, they are considered to be the same person. In the current face recognition data set, the recognition accuracy is 99.38% when the threshold is set to 0.6, but for children and Asians, the recognition rate of the threshold setting of 0.6 is not high, so the threshold is appropriately lowered to 0.4.

[0030] Table 2 Relationship between threshold and recognition accuracy S4: extract moving targets and track trajectories, compare the angles of the starting and ending points of the moving trajectories with the standard values ​​of the moving trajectories, and use face detection and tracking algorithm technology to obtain the location information of the entry boundary of the tourist entering the passenger flow statistics area and the location information of the exit boundary of the tourist leaving the passenger flow statistics area, as described in detail below.

[0031] Extract moving targets and track trajectories. When a tourist enters the entry boundary of the passenger flow statistics area, the tourist portrait is identified through face detection technology, and the feature vector of the face is calculated and saved. The portrait of the entry boundary is taken as the first frame, and the tourist's position is recorded as P0. Next, the tourist's position in each frame is obtained and recorded in the set area until the tourist leaves the exit boundary of the passenger flow statistics area and is recorded as the final departure position Pn. The tourist's position in each frame is connected in sequence to obtain the tourist's travel direction in the passenger flow statistics area.

[0032] exist Figure 7 In the figure, P0 is the point where tourists enter the passenger flow statistics area, and Pn is the point where tourists leave the passenger flow statistics area. Among them, p1, p2, p3, and p4 are the points where tourists are located in each frame in the passenger flow statistics area. The above points are connected in sequence to obtain the approximate movement trajectory of tourists in the passenger flow statistics area. In this part, trajectory tracking only assists in determining the direction of travel, and mainly records the point information of tourists entering the passenger flow statistics area and leaving the passenger flow statistics area. Combined with step S2, the following describes in detail the method of determining whether it is a valid entry or exit according to the movement trajectory. 1. Figure 2 Take the scenic spot shown in the figure as an example. Figure 8 , the way to determine effective entry is: when a tourist enters the passenger flow statistics area from the entry boundary, the point information about the tourist on the entry boundary is recorded as P0, and when a tourist leaves the passenger flow statistics area from the exit boundary, the point information about the tourist on the exit boundary is recorded as Pn. According to step S2, the angle value of the passenger flow statistics area and the angle threshold of the passenger flow statistics trend are used to calculate the angle of the tourist entry trajectory ∠pen as 23 o , that is, it meets two conditions, (1) Visitors enter the area through the entry boundary and leave the area through the exit boundary.

[0033] (2) The angle ∠pen drawn by the straight line between the passenger flow entering and leaving the statistical area and the oblique angle value of the statistical area is 23 o , the value range of the angle threshold ∠Rp in the passenger flow statistics area is 10 to 30, and the above value range can be adjusted according to the location of the camera equipment in the scenic area; 2. If a tourist does not enter the passenger flow statistics area from the entry boundary, does not leave the passenger flow statistics area from the exit boundary, or the angle ∠pen formed by the straight line between the tourist flow statistics area and the oblique angle value of the statistics area is not within the angle threshold range of the passenger flow statistics area, it cannot be counted as a valid entry and will not be counted in the effective passenger flow.

[0034] S5: Comprehensively judge the entry and exit situation based on the feature vector and the movement trajectory, firstly perform deduplication judgment based on the saved facial feature vector of the tourist; 1. Obtain the facial feature vector of the tourists who entered the passenger flow statistics area on that day, and compare it with the facial feature vector saved on that day. If the Euclidean vector is greater than the threshold sim, it is considered that the same person does not need to be counted. If the Euclidean vector is less than the threshold sim, it is considered that the tourist did not enter the park on that day.

[0035] 2. According to step S4, the intersection angle ∠pen determined by the visitor's entry and exit location information and the statistical area angle threshold ∠Rp are used for judgment, and the movement trajectory with ∠pen less than or equal to ∠Rp is taken.

[0036] 3. When both of the above conditions are met, the tourist is considered to have successfully entered or left the park; otherwise, the tourist’s trajectory is considered abnormal and is not counted in the passenger flow.

[0037] S6: Count the passenger flow of the current device for effective entry and exit. According to the above steps, the passenger flow data is counted according to different camera points. For the scenic spot, the entrance and exit cameras are located in different positions. Information is provided for the position and the scenic spot, and the entrance and exit of the surveillance camera are distinguished. Finally, the calculation formula for the real-time number of people in the park E is obtained as follows: In the formula, En is the passenger flow counted by a single entrance surveillance video, and Eg is the passenger flow counted by a single exit surveillance video. The passenger flow counted by all selected entrance surveillance devices and all exit surveillance devices is calculated, and the result is the real-time number of people in the park.

[0038] against Figure 2 The scenic spot shown has two entrances and one exit. According to the comprehensive statistics of steps S1 to S5, at a certain time in the afternoon, the passenger flow entering the two entrances was 492 and 87 respectively, and the passenger flow at the exit was 213. Therefore, the real-time number of people in the park is E = (492 + 87) - 213 = 366.

[0039] S7: Calculate the saturation of the scenic spot based on the passenger flow statistics and output it. According to the maximum carrying capacity determined by the scenic spot under the guidance of the "Guidelines for Determining the Maximum Carrying Capacity of Scenic Spots", the passenger flow entering and leaving the scenic spot is counted by the multi-channel surveillance video of the scenic spot. The formula for determining the passenger flow saturation of the scenic spot is: R = E / (Max SC – k)*100 In the formula: R is the tourist flow saturation of the scenic spot; k is a scenic spot staff member and needs to be eliminated from the scenic spot tourists.

[0040] E is the real-time number of people in the scenic area; Max SC It is the maximum carrying capacity of the scenic area.

[0041] against Figure 2 The maximum carrying capacity of the scenic spot is 6,000 people. The staff of the scenic spot are not sure, but the average daily number is about 120 people. The final calculation is that the passenger flow saturation of the scenic spot at a certain time in the afternoon is: R = 366 / ( 6000 – 120 ) * 100% = 6.22%.

[0042] Glossary Passenger flow statistics area: For single-channel surveillance video equipment, a rectangular area that completely covers the entrance and exit of the scenic area is selected in the video image area, and only tourists passing through this area are analyzed and identified.

[0043] Access boundary: Select an edge in the rectangle of the passenger flow statistics area where tourists enter. When tourists enter the passenger flow statistics area from this edge, face recognition and trajectory tracking are performed, and the location information of tourists entering the passenger flow statistics area is recorded.

[0044] Exit boundary: Select an edge from the rectangle of the passenger flow statistics area where tourists leave. When tourists leave the passenger flow statistics area from this edge, face recognition and trajectory tracking are stopped, and the location information of tourists leaving the passenger flow statistics area is recorded.

[0045] Statistical area oblique angle value: There are many situations in the demarcation of passenger flow statistical areas, so the demarcated rectangle will be rotated at different angles through the intersection of two diagonals. The angle between the image horizontal line and the vertical line of the entry and exit boundaries is set as the statistical area oblique angle value.

[0046] Angle threshold of statistical area: Select any point from the entry boundary of the passenger flow statistical area rectangle, and the angle formed by the line connecting this point to any point on the exit boundary and the perpendicular line of the entry boundary at that point about the oblique angle value of the statistical area.

Claims

1. An intelligent passenger flow saturation analysis method based on multi-channel surveillance videos of scenic spots is characterized by The steps include: S1. Install monitoring equipment at the entrance and exit of the scenic area or select installed monitoring equipment to record monitoring video in real time; S2. Set the passenger flow statistics area, calculate and enter the standard value of the moving trajectory:

1. Set the passenger flow statistics area. Select an area in the entire screen area of ​​the surveillance video recorded in S1 as the passenger flow statistics area. The passenger flow statistics area consists of the travel channel, entry boundary and exit boundary. The monitoring equipment records the entire process of tourists entering the entry boundary and leaving the exit boundary.

2. Calculate and enter the standard value of the moving trajectory, which is composed of the oblique angle value of the statistical area and the angle threshold of the statistical area; (1) Statistical area oblique angle value: select any point P from the entry boundary line in the direction of tourists entering, draw a straight line L1 about point P according to the horizontal angle of the video screen, and then draw a straight line L2 perpendicular to the entry boundary line. The oblique angle value of the statistical area is ∠eng. This oblique angle value represents the general direction of tourists in the entire system. (2) Angle threshold of the statistical area: Draw two diagonal lines in the designated passenger flow statistical rectangular area. The angles between the two diagonal lines and the travel channel are ∠Le and ∠Re respectively. The angle threshold of the statistical area is ∠Rp. Set ∠Rp=∠Le=∠Re. In the recorded surveillance video, P0 is the point where tourists enter the passenger flow statistics area; Pn is the point where tourists leave the passenger flow statistics area; draw a horizontal line L1 about point P0 based on the video screen; the horizontal line L1 intersects with the vertical line L2 of the rectangular area about the entry boundary, and the intersection angle is the oblique angle value ∠eng of the statistical area; Connect points P0 and Pn and extend them to get line segment L3, and record the angle of intersection of L3 and L2 as ∠pen. Compare the value range of ∠pen with the angle threshold ∠Rp of the statistical area; S3, face recognition and recording of face feature vectors, and removal of duplicate passenger flows by feature vector comparison. First, face detection technology is used to obtain feature vectors corresponding to all faces in the video stream, and the Euclidean distance between faces is set. If the Euclidean distance between two faces is less than the set threshold sim, they are considered to be the same person. S4: extract moving targets and track trajectories, compare the angles of the starting and ending points of the moving trajectories with the standard values ​​of the moving trajectories, and use face detection and tracking algorithm technology to obtain the location information of the tourists entering the entry boundary of the passenger flow statistical area and the location information of the tourists leaving the exit boundary of the passenger flow statistical area. Extract moving targets and track trajectories. When tourists enter the entry boundary of the passenger flow statistical area, use face detection technology to identify the portrait of the tourists, calculate the feature vector of the face and save it. In addition, the portrait of the entry boundary is taken as the first frame, and the position of the tourists is recorded as P0. Next, the position of the tourists in each frame is obtained and recorded in the set area until the tourists leave the exit boundary of the passenger flow statistical area and record it as the final exit position Pn. The positions of the tourists in each frame are connected in sequence to obtain the direction of travel of the tourists in the passenger flow statistical area. S5: Comprehensively judge the entry and exit situation based on the feature vector and the movement trajectory, firstly perform deduplication judgment based on the saved facial feature vector of the tourist; 1. Obtain the facial feature vector of the tourist who enters the passenger flow statistics area on that day, and compare it with the facial feature vector saved on that day. If the Euclidean vector is greater than the threshold sim, it is considered that the same person does not need to be counted; if the Euclidean vector is less than the threshold sim, it is considered that the tourist did not enter the park on that day; 2. According to step S4, the intersection angle ∠pen determined by the visitor's entry and exit location information and the statistical area angle threshold ∠Rp are used for judgment, and the action trajectory with ∠pen less than or equal to ∠Rp is selected; 3. If both of the above conditions are met, the visitor is considered to have successfully entered or left the park. Otherwise, the visitor's trajectory is considered abnormal and will not be counted in the visitor flow. S6: Count the passenger flow of the current device for effective entry and exit. According to the above steps, the passenger flow data is counted according to different camera points. For the scenic spot, the entrance and exit cameras are located in different positions. Information is provided for the position and the scenic spot, and the entrance and exit of the surveillance camera are distinguished. Finally, the calculation formula for the real-time number of people in the park E is obtained as follows: In the formula, En is the passenger flow counted by a single entrance surveillance video, and Eg is the passenger flow counted by a single exit surveillance video. The passenger flow counted by all selected entrance surveillance devices and all exit surveillance devices is calculated, and the result is the real-time number of people in the park; S7: Calculate the saturation of the scenic spot based on the passenger flow statistics and output it. According to the maximum carrying capacity determined by the scenic spot and the passenger flow entering and leaving the scenic spot counted by the multi-channel surveillance video of the scenic spot, the passenger flow saturation formula of the scenic spot can be finally determined as follows: R = E / (MaxSC – k) * 100 In the formula: R is the tourist flow saturation of the scenic spot; k is the staff of the scenic spot and needs to be eliminated from the tourists in the scenic spot; E is the real-time number of people in the scenic area; MaxSC is the maximum carrying capacity of the scenic area.

2. According to claim 1, the intelligent passenger flow saturation analysis method based on multi-channel surveillance video of scenic spots is characterized in that In S1, monitoring equipment is installed at the entrance and exit of the scenic area or the installed monitoring equipment is selected. The specific technical requirements of the monitoring equipment are as follows: The resolution of the monitoring equipment is above 1080P; the monitoring equipment is located above the ticket gate, queuing area or entrance and exit of the scenic area, at a height of 2.5-3.5 meters, and records the monitoring video in real time.

3. According to claim 1, the intelligent passenger flow saturation analysis method based on multi-channel surveillance video of scenic spots is characterized in that The statistical area angle threshold ∠Rp in S2 ranges from 10 to 30°.

Images