Visitor flow early warning method and system for smart scenic area

Abstract

The invention relates to the technical field of intelligent tourism, and discloses a pedestrian flow early warning method and system for an intelligent scenic spot, and the system comprises a scenic spot depth image collection module, a depth image preprocessing module, a pedestrian target detection module, a pedestrian target tracking module, a risk assessment module, and a partition early warning regulation and control module. Depth images of all monitoring sub-areas of a smart scenic spot are collected, a high-precision three-dimensional dynamic scene model is generated in combination with a laser radar, the depth images are preprocessed based on a bilateral filtering algorithm, and pedestrian targets of all the monitoring sub-areas of the smart scenic spot are analyzed and calibrated through circularity calculation and circularity size constraint. The motion trail of the pedestrian target is tracked by using the gravity center trail of the human head area, the risk assessment of the pedestrian flow of each monitoring sub-area of the smart scenic spot is completed, the pedestrian flow early warning information is published in real time, a scientific basis is provided for scenic spot management, and the tour experience of tourists is optimized.

Description

Technical Field

[0001] The present invention relates to the field of smart tourism technology, and more specifically to a method and system for early warning of crowd flow in a smart scenic area. Background Art

[0002] With the booming tourism industry, the number of visitors to scenic spots has increased dramatically, especially during holidays and peak tourist seasons, when popular scenic spots often face overcrowding. This overcrowding not only diminishes the visitor experience but can also pose safety risks, such as stampedes and resource overload. Traditional crowd management methods rely on manual statistics or simple gate counting, which suffer from data lag, limited coverage, and slow response times, making them unable to meet the demands of refined management in modern scenic spots. Against this backdrop, smart scenic spot crowd warning technology has emerged, aiming to achieve real-time monitoring, accurate prediction, and dynamic regulation through intelligent means, providing technical support for safe and efficient scenic spot operations.

[0003] Existing technologies, such as IoT-based sensor networks or Wi-Fi probes, can roughly locate pedestrians using their mobile phone signals, but are susceptible to device density and signal interference. While deep learning methods based on computer vision can identify crowd density using cameras, they are limited by lighting, occlusion, and false detection in high-density scenarios. When pedestrians obscure the area, the algorithm cannot segment individuals, resulting in small or dynamic targets going undetected. Furthermore, existing early warning systems focus primarily on real-time monitoring and lack adaptive control capabilities. Therefore, how to calibrate the dynamic movement of target pedestrians and build a real-time, accurate, and scalable crowd flow early warning system has become a key challenge in the development of smart scenic spots. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a pedestrian flow warning system for a smart scenic spot to solve the problems existing in the above-mentioned background technology.

[0005] The present invention provides the following technical solution: a pedestrian flow warning system for a smart scenic area, comprising: a scenic area depth image acquisition module, a depth image preprocessing module, a pedestrian target detection module, a pedestrian target tracking module, a risk assessment module, and a zoning warning control module; The scenic area depth image acquisition module uses a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combines it with a lidar to generate a high-precision three-dimensional dynamic scene model; The depth image preprocessing module preprocesses the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after preprocessing; The pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. It calibrates pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis. The pedestrian target tracking module includes a target feature extraction unit and a motion trajectory tracking unit, which uses the center of gravity trajectory of the human head area to track the motion trajectory of the pedestrian target and complete dynamic monitoring of the pedestrian target; The risk assessment module performs risk assessment on the pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets; The zoning early warning and control module includes an early warning information push unit and an intelligent control path planning unit, which completes the zoning early warning and control of each monitoring sub-area of ​​the smart scenic area based on the risk assessment results.

[0006] Preferably, the specific contents of the scenic area depth image acquisition module are as follows: Divide the monitoring area of ​​the smart scenic area into n monitoring sub-areas, i = 1, 2, 3, ..., n, where i represents the number of each monitoring sub-area; Depth cameras are installed in each monitoring sub-area to collect depth images of each monitoring sub-area in the smart scenic area. The components of the depth camera equipment include: a microphone array, an infrared transmitter, an infrared receiver, a base motor, and an RGB camera. The infrared transmitter and infrared receiver acquire depth images by transmitting and receiving infrared rays, and the RGB camera is used to acquire color images. The microphone array uses four-element linear microphone array technology to provide full coverage of each monitoring sub-area in the smart scenic area, and the base motor has an automatic focus tracking function. LiDAR is covered in each monitoring sub-area to obtain the terrain data of each monitoring sub-area of ​​the smart scenic area, and then integrated through the geographic information system to generate a high-precision three-dimensional dynamic scene model.

[0007] Preferably, the depth image preprocessing module preprocesses the depth images of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm, extracts the original pixel values ​​of the depth images of each monitoring sub-area, performs weight processing in the spatial domain and range domain on the original pixel values ​​of the depth images of each monitoring sub-area, and obtains the image pixel values ​​after preprocessing the depth images of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm.

[0008] Preferably, the pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. The specific contents are as follows: The depth image segmentation unit is configured to evenly segment the pre-processed depth image into four regions, wherein the pixel value of each region is one fourth of the pixel value of the pre-processed depth image; The maximum stable extreme value region determining unit: the gray value threshold of each region image changes from the minimum value to the maximum value m times, generates m binary images, and extracts the connected regions in each binary image; Perform regional detection based on the connected regions in each binary image: Bright region detection: The threshold increases from low to high, and the threshold at which each pixel is first included in the connected region is recorded; Dark region detection: The threshold decreases from high to low, and the threshold at which each pixel is first included in the connected region is recorded; Generate an extreme value region sequence for each regional image based on the regional detection results, perform stability judgment on the extreme value region sequence for each regional image, and screen out the maximum stable extreme value region for each regional image based on the stability judgment formula; The maximum stable extreme value region merging unit is configured to merge the maximum stable extreme value regions of each screened regional image to obtain the pre-processed pedestrian candidate regions of each monitoring sub-region of the smart scenic area; The human head region acquisition unit performs circularity analysis based on the area and perimeter of the pedestrian candidate regions in each monitoring sub-region, calculates the circularity of each candidate region, and calibrates the pedestrian targets in each monitoring sub-region of the smart scenic area based on the circularity size constraint.

[0009] Preferably, the pedestrian target tracking module includes a target feature extraction unit and a motion trajectory tracking unit, and the specific contents are as follows: The target feature extraction unit extracts features of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, obtains the head contours of pedestrians in each monitoring sub-area of ​​the smart scenic area, and extracts the centroid coordinates of the pedestrian head contours. The centroid coordinates are two-dimensional coordinate points, expressed as ; The motion trajectory tracking unit counts the edge points of the pedestrian head contours in each monitoring sub-area of ​​the smart scenic area according to the pedestrian head contours obtained by the target feature extraction unit, and analyzes the distance from each edge point of the pedestrian head contour to the centroid coordinates of the pedestrian head contour , where j represents the number of edge points of the pedestrian head contour in each monitoring sub-area, and the maximum distance from the edge point to the center of gravity coordinate is taken as the circle radius. The expression is: , and tracking the movement trajectory of the pedestrian target's head changes based on the circular area formed by the center of gravity coordinates and the circle radius; The pedestrian flow is updated in real time based on the movement trajectory of pedestrian targets in each monitoring sub-area, and the pedestrian flow is increased or decreased based on the preset center of gravity movement direction: when the movement trajectory of a pedestrian target is detected entering a monitoring sub-area, it is judged as an entry state, and the pedestrian flow in the corresponding monitoring sub-area is increased by one; when the movement trajectory of a pedestrian target is detected leaving a monitoring sub-area, it is judged as a departure state, and the pedestrian flow in the corresponding monitoring sub-area is decreased by one.

[0010] Preferably, the risk assessment module performs risk assessment on the pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets. The specific contents are as follows: Receive the number of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, receive the tracking results of the pedestrian target movement trajectories in each monitoring sub-area of ​​the smart scenic area by the pedestrian target tracking module, and obtain the dynamic pedestrian flow in each monitoring sub-area of ​​the smart scenic area; Risk assessment is performed based on the dynamic flow of people in each monitoring sub-area of ​​the smart scenic area: when the dynamic flow of people in the monitoring sub-area is greater than or equal to the preset threshold, the risk assessment result is high risk; when the dynamic flow of people in the monitoring sub-area is less than the preset threshold, the risk assessment result is low risk.

[0011] Preferably, the partition warning control module includes the following specific contents: a warning information push unit and a control path intelligent planning unit: The warning information push unit receives the risk assessment result and sends the warning information to the human-computer interaction terminal when the risk assessment result is high risk. The human-computer interaction terminal is connected to the display device and audio device of each monitoring sub-area. The human-computer interaction terminal broadcasts the crowd evacuation of the high-risk monitoring sub-area through the audio device and activates the control path intelligent planning unit; The control path intelligent planning unit receives the start instruction of the risk assessment result, combines the scenic area depth image acquisition module to generate a high-precision three-dimensional dynamic scene model to perform intelligent planning of the control path, and intelligently displays it in the high-risk area through the display device.

[0012] A method for early warning of crowd flow in a smart scenic area comprises the following steps: Step S01: Use a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combine it with a lidar to generate a high-precision three-dimensional dynamic scene model; Step S02: pre-processing the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after pre-processing; Step S03: calibrate pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis; Step S04: Tracking the motion trajectory of the pedestrian target using the center of gravity trajectory of the human head area to complete dynamic monitoring of the pedestrian target; Step S05: Based on the calibrated pedestrian targets in each monitoring sub-area of ​​the smart scenic area and the results of tracking the pedestrian targets' motion trajectories, a risk assessment of the pedestrian flow in each monitoring sub-area of ​​the smart scenic area is performed; Step S06: Based on the risk assessment results, complete the zoning early warning control of each monitoring sub-area of ​​the smart scenic area.

[0013] Technical effects and advantages of the present invention: The present invention solves the problems of missed detection and false detection in existing technologies due to limitations of lighting, occlusion and high-density scenes by providing a scenic area depth image acquisition module, a depth image preprocessing module, a pedestrian target detection module, a pedestrian target tracking module, a risk assessment module and a zoning warning and control module. Through joint weighting of the spatial domain and range domain, the edge information of the depth image is retained while the noise information of the depth image is reduced; the head is accurately calibrated through regional area and perimeter constraints to avoid missed detections and false detections caused by limb occlusion or dense crowds in full-body detection; the motion trajectory of the center of gravity in the head area is used to achieve high-precision dynamic tracking of pedestrian targets, effectively improving the detection accuracy and tracking stability of pedestrian targets in complex scenarios. Furthermore, combined with a high-precision three-dimensional dynamic scene model, it can accurately assess the crowd flow risk of each monitoring sub-area in real time and trigger early warning and control mechanisms in a timely manner, thereby significantly enhancing the crowd flow management capabilities of smart scenic spots and providing tourists with a safer and more comfortable sightseeing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a structural diagram of a pedestrian flow warning system for a smart scenic area.

[0015] Figure 2 The figure is a flow chart of a method for early warning of crowd flow in a smart scenic spot. DETAILED DESCRIPTION

[0016] The technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely examples. The method and system for early warning of pedestrian flow in a smart scenic area involved in the present invention are not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, the present invention provides a pedestrian flow warning system for a smart scenic area, comprising: a scenic area depth image acquisition module, a depth image preprocessing module, a pedestrian target detection module, a pedestrian target tracking module, a risk assessment module, and a zoning warning control module; The scenic area depth image acquisition module uses a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combines it with a lidar to generate a high-precision three-dimensional dynamic scene model; The depth image preprocessing module preprocesses the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after preprocessing; The pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. It calibrates pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis. The pedestrian target tracking module includes a target feature extraction unit and a motion trajectory tracking unit, which uses the center of gravity trajectory of the human head area to track the motion trajectory of the pedestrian target and complete dynamic monitoring of the pedestrian target; The risk assessment module performs risk assessment on the pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets; The zoning early warning and control module includes an early warning information push unit and an intelligent control path planning unit, which completes the zoning early warning and control of each monitoring sub-area of ​​the smart scenic area based on the risk assessment results.

[0018] In this embodiment, it should be specifically explained that the specific contents of the scenic area depth image acquisition module are as follows: Divide the monitoring area of ​​the smart scenic area into n monitoring sub-areas, i = 1, 2, 3, ..., n, where i represents the number of each monitoring sub-area; Depth cameras are installed in each monitoring sub-area to collect depth images of each monitoring sub-area in the smart scenic area. The components of the depth camera equipment include: a microphone array, an infrared transmitter, an infrared receiver, a base motor, and an RGB camera. The infrared transmitter and infrared receiver acquire depth images by transmitting and receiving infrared rays, and the RGB camera is used to acquire color images. The microphone array uses four-element linear microphone array technology to provide full coverage of each monitoring sub-area in the smart scenic area, and the base motor has an automatic focus tracking function. LiDAR is covered in each monitoring sub-area to obtain the terrain data of each monitoring sub-area of ​​the smart scenic area. After integration through the geographic information system, a high-precision three-dimensional dynamic scene model is generated. The three-dimensional dynamic scene model is used to control the path intelligent planning unit to perform intelligent path planning.

[0019] In this embodiment, it should be specifically explained that the depth image preprocessing module preprocesses the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to eliminate the impact of illumination changes and occlusion on the data, extracts the original pixel value of the depth image of each monitoring sub-area, performs weight processing in the spatial domain and range domain on the original pixel value of the depth image of each monitoring sub-area, and obtains the image pixel value after the depth image of each monitoring sub-area of ​​the smart scenic area is preprocessed based on the bilateral filtering algorithm. The calculation formula of the bilateral filtering algorithm is: ,in The image represents the pre-processed depth image of each monitoring sub-area of ​​the smart scenic spot. The pixel value at The depth image of each monitoring sub-area is represented by The original pixel value, Represents the spatial domain weight of the original pixel value of the depth image of each monitoring sub-area, Represents the range domain weight of the original pixel value of the depth image of each monitoring sub-area.

[0020] In this embodiment, it should be specifically explained that the pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. The specific contents are as follows: The depth image segmentation unit is configured to evenly segment the pre-processed depth image into four regions, wherein the pixel value of each region is one fourth of the pixel value of the pre-processed depth image; The maximum stable extreme value region determining unit: the gray value threshold of each region image changes from the minimum value to the maximum value m times, generates m binary images, and extracts the connected regions in each binary image; Perform regional detection based on the connected regions in each binary image: Bright region detection: The threshold increases from low to high, and the threshold at which each pixel is first included in the connected region is recorded; Dark region detection: The threshold decreases from high to low, and the threshold at which each pixel is first included in the connected region is recorded; Generate an extreme value region sequence for each regional image based on the regional detection results, perform stability judgment on the extreme value region sequence for each regional image, and screen out the maximum stable extreme value region for each regional image based on the stability judgment formula; The stability judgment formula is: ,in represents the stability analysis results of the extreme value region sequence, Indicates area The pixel area, Indicates the threshold change step size. Take the local minimum at the threshold t, and When it is less than the preset stability threshold, is the maximum stable extreme value area; The maximum stable extreme value region merging unit is configured to merge the maximum stable extreme value regions of each screened regional image to obtain the pre-processed pedestrian candidate regions of each monitoring sub-region of the smart scenic area; The human head region acquisition unit performs circularity analysis based on the area and perimeter of the pedestrian candidate regions in each monitoring sub-region, calculates the circularity of each candidate region, and calibrates the pedestrian targets in each monitoring sub-region of the smart scenic area based on the circularity size constraint; The calculation formula for the circularity of each candidate area is: ,in represents the circularity of each candidate region, represents the area of ​​the pedestrian candidate region in each monitoring sub-region, The perimeter of the pedestrian candidate area in each monitoring sub-area; The circularity of each candidate area is compared with the preset circularity judgment threshold range. If the circularity of the candidate area is within the preset circularity judgment threshold range, pedestrian target calibration is performed. If the circularity of the candidate area is not within the preset circularity judgment threshold range, pedestrian target calibration is not performed.

[0021] In this embodiment, it should be specifically explained that the pedestrian target tracking module includes the target feature extraction unit and the motion trajectory tracking unit. The specific contents are as follows: The target feature extraction unit extracts features of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, obtains the head contours of pedestrians in each monitoring sub-area of ​​the smart scenic area, and extracts the centroid coordinates of the pedestrian head contours. The centroid coordinates are two-dimensional coordinate points, expressed as ; The motion trajectory tracking unit counts the edge points of the pedestrian head contours in each monitoring sub-area of ​​the smart scenic area according to the pedestrian head contours obtained by the target feature extraction unit, and analyzes the distance from each edge point of the pedestrian head contour to the centroid coordinates of the pedestrian head contour , where j represents the number of edge points of the pedestrian head contour in each monitoring sub-area, and the maximum distance from the edge point to the center of gravity coordinate is taken as the circle radius. The expression is: , and tracking the movement trajectory of the pedestrian target's head changes based on the circular area formed by the center of gravity coordinates and the circle radius; When the count result of edge points in the i-th monitoring sub-area is 6, and the distance between the 6 edge points and the centroid coordinates of the pedestrian head contour is They are: 10cm, 8cm, 7cm, 7cm, 9cm and 8cm respectively. ; The pedestrian flow is updated in real time based on the movement trajectory of pedestrian targets in each monitoring sub-area, and the pedestrian flow is increased or decreased based on the preset center of gravity movement direction: when the movement trajectory of a pedestrian target is detected entering a monitoring sub-area, it is judged as an entry state, and the pedestrian flow in the corresponding monitoring sub-area is increased by one; when the movement trajectory of a pedestrian target is detected leaving a monitoring sub-area, it is judged as a departure state, and the pedestrian flow in the corresponding monitoring sub-area is decreased by one.

[0022] In this embodiment, it should be specifically explained that the risk assessment module performs risk assessment of pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets. The specific contents are as follows: Receive the number of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, receive the tracking results of the pedestrian target movement trajectories in each monitoring sub-area of ​​the smart scenic area by the pedestrian target tracking module, and obtain the dynamic pedestrian flow in each monitoring sub-area of ​​the smart scenic area; Risk assessment is performed based on the dynamic flow of people in each monitoring sub-area of ​​the smart scenic area: when the dynamic flow of people in the monitoring sub-area is greater than or equal to the preset threshold, the risk assessment result is high risk; when the dynamic flow of people in the monitoring sub-area is less than the preset threshold, the risk assessment result is low risk.

[0023] In this embodiment, it should be specifically explained that the specific contents of the zone warning control module, including the warning information push unit and the control path intelligent planning unit, are as follows: The warning information push unit receives the risk assessment result and sends the warning information to the human-computer interaction terminal when the risk assessment result is high risk. The human-computer interaction terminal is connected to the display device and audio device of each monitoring sub-area. The human-computer interaction terminal broadcasts the crowd evacuation of the high-risk monitoring sub-area through the audio device and activates the control path intelligent planning unit; The control path intelligent planning unit receives the start instruction of the risk assessment result, combines the scenic area depth image acquisition module to generate a high-precision three-dimensional dynamic scene model to perform intelligent planning of the control path, and intelligently displays it in the high-risk area through the display device.

[0024] like Figure 2 As shown, in this embodiment, it should be specifically explained that a method for early warning of pedestrian flow in a smart scenic area includes the following steps: Step S01: Use a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combine it with a lidar to generate a high-precision three-dimensional dynamic scene model; Step S02: pre-processing the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after pre-processing; Step S03: calibrate pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis; Step S04: Tracking the motion trajectory of the pedestrian target using the center of gravity trajectory of the human head area to complete dynamic monitoring of the pedestrian target; Step S05: Based on the calibrated pedestrian targets in each monitoring sub-area of ​​the smart scenic area and the results of tracking the pedestrian targets' motion trajectories, a risk assessment of the pedestrian flow in each monitoring sub-area of ​​the smart scenic area is performed; Step S06: Based on the risk assessment results, complete the zoning early warning control of each monitoring sub-area of ​​the smart scenic area.

[0025] In this embodiment, it should be specifically explained that the difference between this embodiment and the prior art lies in that this embodiment is provided with a scenic area depth image acquisition module, a depth image preprocessing module, a pedestrian target detection module, a pedestrian target tracking module, a risk assessment module, and a zoning warning and control module, thereby solving the problems of missed detection and false detection in the prior art due to limitations of lighting, occlusion, and high-density scenes. Through joint weighting of the spatial domain and range domain, the edge information of the depth image is retained while the noise information of the depth image is reduced; the head is accurately calibrated through regional area and perimeter constraints to avoid missed detections and false detections caused by limb occlusion or dense crowds in full-body detection; the motion trajectory of the center of gravity in the head area is used to achieve high-precision dynamic tracking of pedestrian targets, effectively improving the detection accuracy and tracking stability of pedestrian targets in complex scenarios. Furthermore, combined with a high-precision three-dimensional dynamic scene model, it can accurately assess the crowd flow risk of each monitoring sub-area in real time and trigger early warning and control mechanisms in a timely manner, thereby significantly enhancing the crowd flow management capabilities of smart scenic spots and providing tourists with a safer and more comfortable sightseeing environment.

[0026] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0027] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A pedestrian flow warning system for a smart scenic area, characterized by: include: Scenic area depth image acquisition module, depth image preprocessing module, pedestrian target detection module, pedestrian target tracking module, risk assessment module and zoning warning and control module; The scenic area depth image acquisition module uses a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combines it with a lidar to generate a high-precision three-dimensional dynamic scene model; The depth image preprocessing module preprocesses the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after preprocessing; The pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. It calibrates pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis. The pedestrian target tracking module includes a target feature extraction unit and a motion trajectory tracking unit, which uses the center of gravity trajectory of the human head area to track the motion trajectory of the pedestrian target and complete dynamic monitoring of the pedestrian target; The risk assessment module performs risk assessment on the pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets; The zoning early warning and control module includes an early warning information push unit and an intelligent control path planning unit, which completes the zoning early warning and control of each monitoring sub-area of ​​the smart scenic area based on the risk assessment results.

2. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The specific contents of the scenic area depth image acquisition module are as follows: Divide the monitoring area of ​​the smart scenic area into n monitoring sub-areas, i = 1, 2, 3, ..., n, where i represents the number of each monitoring sub-area; Depth cameras are installed in each monitoring sub-area to collect depth images of each monitoring sub-area in the smart scenic area. The components of the depth camera equipment include: a microphone array, an infrared transmitter, an infrared receiver, a base motor, and an RGB camera. The infrared transmitter and infrared receiver acquire depth images by transmitting and receiving infrared rays, and the RGB camera is used to acquire color images. The microphone array uses four-element linear microphone array technology to provide full coverage of each monitoring sub-area in the smart scenic area, and the base motor has an automatic focus tracking function. LiDAR is covered in each monitoring sub-area to obtain the terrain data of each monitoring sub-area of ​​the smart scenic area, and then integrated through the geographic information system to generate a high-precision three-dimensional dynamic scene model.

3. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The depth image preprocessing module preprocesses the depth images of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm, extracts the original pixel values ​​of the depth images of each monitoring sub-area, performs weight processing in the spatial domain and range domain on the original pixel values ​​of the depth images of each monitoring sub-area, and obtains the image pixel values ​​after the depth images of each monitoring sub-area of ​​the smart scenic area are preprocessed based on the bilateral filtering algorithm.

4. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The pedestrian target detection module includes a depth image segmentation unit, a maximum stable extreme value region determination unit, a maximum stable extreme value region merging unit, and a human head region acquisition unit. The specific contents are as follows: The depth image segmentation unit is configured to evenly segment the pre-processed depth image into four regions, wherein the pixel value of each region is one fourth of the pixel value of the pre-processed depth image; The maximum stable extreme value region determining unit: the gray value threshold of each region image changes from the minimum value to the maximum value m times, generates m binary images, and extracts the connected regions in each binary image; Perform regional detection based on the connected regions in each binary image: Bright region detection: The threshold increases from low to high, and the threshold at which each pixel is first included in the connected region is recorded; Dark region detection: The threshold decreases from high to low, and the threshold at which each pixel is first included in the connected region is recorded; Generate an extreme value region sequence for each regional image based on the regional detection results, perform stability judgment on the extreme value region sequence for each regional image, and screen out the maximum stable extreme value region for each regional image based on the stability judgment formula; The maximum stable extreme value region merging unit is configured to merge the maximum stable extreme value regions of each screened regional image to obtain the pre-processed pedestrian candidate regions of each monitoring sub-region of the smart scenic area; The human head region acquisition unit performs circularity analysis based on the area and perimeter of the pedestrian candidate regions in each monitoring sub-region, calculates the circularity of each candidate region, and calibrates the pedestrian targets in each monitoring sub-region of the smart scenic area based on the circularity size constraint.

5. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The pedestrian target tracking module includes a target feature extraction unit and a motion trajectory tracking unit. The specific contents are as follows: The target feature extraction unit extracts features of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, obtains the head contours of pedestrians in each monitoring sub-area of ​​the smart scenic area, and extracts the centroid coordinates of the pedestrian head contours. The centroid coordinates are two-dimensional coordinate points, expressed as ; The motion trajectory tracking unit counts the edge points of the pedestrian head contours in each monitoring sub-area of ​​the smart scenic area according to the pedestrian head contours obtained by the target feature extraction unit, and analyzes the distance from each edge point of the pedestrian head contour to the centroid coordinates of the pedestrian head contour , where j represents the number of edge points of the pedestrian head contour in each monitoring sub-area, and the maximum distance from the edge point to the center of gravity coordinate is taken as the circle radius. The expression is: , and tracking the movement trajectory of the pedestrian target's head changes based on the circular area formed by the center of gravity coordinates and the circle radius; The pedestrian flow is updated in real time based on the movement trajectory of pedestrian targets in each monitoring sub-area, and the pedestrian flow is increased or decreased based on the preset center of gravity movement direction: when the movement trajectory of a pedestrian target is detected entering a monitoring sub-area, it is judged as an entry state, and the pedestrian flow in the corresponding monitoring sub-area is increased by one; when the movement trajectory of a pedestrian target is detected leaving a monitoring sub-area, it is judged as a departure state, and the pedestrian flow in the corresponding monitoring sub-area is decreased by one.

6. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The risk assessment module performs risk assessment on the pedestrian flow in each monitoring sub-area of ​​the smart scenic area based on the pedestrian targets calibrated by the pedestrian target detection module and the results of the pedestrian target tracking module tracking the movement trajectory of the pedestrian targets. The specific contents are as follows: Receive the number of pedestrian targets in each monitoring sub-area of ​​the smart scenic area calibrated by the pedestrian target detection module, receive the tracking results of the pedestrian target movement trajectories in each monitoring sub-area of ​​the smart scenic area by the pedestrian target tracking module, and obtain the dynamic pedestrian flow in each monitoring sub-area of ​​the smart scenic area; Risk assessment is performed based on the dynamic flow of people in each monitoring sub-area of ​​the smart scenic area: when the dynamic flow of people in the monitoring sub-area is greater than or equal to the preset threshold, the risk assessment result is high risk; when the dynamic flow of people in the monitoring sub-area is less than the preset threshold, the risk assessment result is low risk.

7. The pedestrian flow warning system for a smart scenic spot according to claim 1 is characterized by: The specific contents of the zone warning and control module, including the warning information push unit and the control path intelligent planning unit, are as follows: The warning information push unit receives the risk assessment result and sends the warning information to the human-computer interaction terminal when the risk assessment result is high risk. The human-computer interaction terminal is connected to the display device and audio device of each monitoring sub-area. The human-computer interaction terminal broadcasts the crowd evacuation of the high-risk monitoring sub-area through the audio device and activates the control path intelligent planning unit; The control path intelligent planning unit receives the start instruction of the risk assessment result, combines the scenic area depth image acquisition module to generate a high-precision three-dimensional dynamic scene model to perform intelligent planning of the control path, and intelligently displays it in the high-risk area through the display device.

8. A method for early warning of pedestrian flow in a smart scenic area, for using the pedestrian flow early warning system of a smart scenic area according to any one of claims 1 to 7, characterized in that: The following steps are involved: Step S01: Use a depth camera device to collect depth images of each monitoring sub-area of ​​the smart scenic area, and combine it with a lidar to generate a high-precision three-dimensional dynamic scene model; Step S02: pre-processing the depth image of each monitoring sub-area of ​​the smart scenic area based on the bilateral filtering algorithm to obtain the depth image data of each monitoring sub-area of ​​the smart scenic area after pre-processing; Step S03: calibrate pedestrian targets in each monitoring sub-area of ​​the smart scenic area through circularity calculation and circularity size constraint analysis; Step S04: Tracking the motion trajectory of the pedestrian target using the center of gravity trajectory of the human head area to complete dynamic monitoring of the pedestrian target; Step S05: Based on the calibrated pedestrian targets in each monitoring sub-area of ​​the smart scenic area and the results of tracking the pedestrian targets' motion trajectories, a risk assessment of the pedestrian flow in each monitoring sub-area of ​​the smart scenic area is performed; Step S06: Based on the risk assessment results, complete the zoning early warning control of each monitoring sub-area of ​​the smart scenic area.

Images