Intelligent campus safety early warning system based on video twinborn behavior analysis
Through video twin behavior analysis technology, combined with cloud computing and infrared feature processes, a security warning network is established, which solves the problems of monitoring blind spots and data analysis errors in campus safety warning, and improves the accuracy and efficiency of campus safety warning.
Patent Information
- Application Number
- CN202410988059.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing campus safety warning technology has problems such as blind spots in monitoring and data analysis results that are prone to errors, resulting in the inability to fully cover campus safety risks and reduce the accuracy and efficiency of security warnings.
Using a smart campus safety warning system based on video twin behavior analysis, the cloud computing platform combines campus scene data acquisition module, early warning event analysis module and security warning module to generate event feature chains and security warning networks, and use infrared feature processes to output early warning information. Improve early warning accuracy and efficiency.
It achieves the accuracy and efficiency of campus safety warning while ensuring the accuracy and efficiency of campus safety warnings, ensuring timely identification and prevention of potential risks.
Smart Images

Figure CN119006238A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of campus safety management, and specifically to a smart campus safety early warning system based on video twin behavior analysis. Background Art
[0002] Campus safety early warning refers to the use of various means and technologies to timely grasp the possible safety risks and emergencies on campus, and take corresponding measures to respond and deal with them. The purpose of campus safety early warning is to protect the life and property safety of teachers and students in the school.
[0003] The existing campus safety early warning technology has the following defects:
[0004] Video surveillance technology: The video surveillance system monitors the activities of students and faculty members by setting up cameras on campus. However, due to the large campus area or the uneven distribution of surveillance equipment, there may be some blind spots that cannot be fully covered, which is prone to monitoring gaps.
[0005] Big data analysis technology: Establish early warning models based on data analysis results to identify potential risks and abnormal behaviors, and discover and prevent risks in advance. However, due to the different conditions on campus at different times, data analysis results are prone to errors, resulting in abnormal early warning results.
[0006] Therefore, how to improve the accuracy and efficiency of campus safety warnings while ensuring the accuracy of obtaining campus status data is a difficulty in existing technology. For this purpose, a smart campus safety warning system based on video twin behavior analysis is provided. Summary of the invention
[0007] In order to solve the above technical problems, the purpose of the present invention is to provide a smart campus safety warning system based on video twin behavior analysis.
[0008] In order to achieve the above object, the present invention provides the following technical solutions:
[0009] A smart campus safety warning system based on video twin behavior analysis includes a cloud computing platform, which is communicatively connected to a campus scene data acquisition module, a warning event analysis module, and a safety warning module;
[0010] The campus scene data collection module is used to set the class time period, break time period and holiday time period, and install a number of cameras with overlapping shooting ranges on the campus, so as to collect the campus scene data set in each time period through the cameras;
[0011] The warning event analysis module is used to generate event process feature images and warning feature images based on a number of pre-stored campus dangerous events, and extract the corresponding feature process from the event process feature images and the warning feature images, and then establish an event feature chain based on the feature process, establish a safety warning network based on the correlation between different event feature chains, and mark the corresponding time period;
[0012] The safety warning module is used to generate corresponding infrared feature processes according to campus scene data in various time periods, input the infrared feature processes into the safety warning network, and then output corresponding safety warning information and generate corresponding warning event decisions.
[0013] Furthermore, the collection process of the campus scene data set includes:
[0014] Upload the semester table to the campus scene data collection module, where the semester table records the daily course names, break times, dismissal times, and vacation times during the corresponding semester;
[0015] Then the campus scene data collection module establishes a semester timeline according to the length of the semester table, and divides the semester timeline into several class periods, break periods, and holiday periods;
[0016] At the same time, the campus scene data acquisition module installs several groups of optical cameras and infrared cameras at various locations on the campus, and sets the same number for the optical cameras and infrared cameras in the same group;
[0017] A data upload cycle is set, wherein the duration of the data upload cycle is less than half of the duration of the break period, and each group of optical cameras and infrared cameras captures optical video data and infrared video data within their shooting ranges at the beginning of each data upload cycle, and sends the optical video data and infrared video data to the campus data acquisition module at the end of the data upload cycle;
[0018] Whenever a data upload cycle ends, the campus data acquisition module marks each group of optical video data and infrared video data with the number of the corresponding optical camera and infrared camera, and integrates all optical video data and infrared video data to generate the corresponding campus scene data set, and marks the corresponding acquisition time.
[0019] Furthermore, the shooting range and shooting angle of each group of optical cameras and infrared cameras remain consistent at any time, and the shooting angle can be moved within a certain range. At the same time, the shooting range and shooting angle of each group of optical cameras and infrared cameras at adjacent spatial positions have an intersection relationship at any time.
[0020] Furthermore, the feature extraction process includes:
[0021] The early warning event analysis module is preset with several campus dangerous events, and the campus dangerous events include event names, time periods, event process feature videos, and warning feature videos;
[0022] An event timeline is established, and several time nodes are set on the event timeline. The early warning event analysis module divides the event process feature video and the warning feature video into several event process feature images and warning feature images by frame;
[0023] The event process characteristic image and the warning characteristic image are grayed out respectively, and the pixel value of each pixel in the event process characteristic image and the warning characteristic image is obtained, and the pixel blur radius r is set, and each pixel in the event process characteristic image and the warning characteristic image is selected as the circle center pixel respectively, and then the pixel value of the circle center pixel and all pixels within the selected pixel blur radius r are averaged;
[0024] Setting a characteristic pixel value threshold, and then comparing the pixel value of each pixel in the event process characteristic image and the warning characteristic image after averaging each pixel with the characteristic pixel value threshold;
[0025] Pixels with values greater than or equal to the characteristic pixel threshold are recorded as characteristic pixels, and no operation is performed on the remaining pixels;
[0026] The feature pixels at adjacent positions are connected in sequence to obtain the feature processes of the feature images of each event process and the feature process of the warning feature images.
[0027] Furthermore, the pixel value averaging formula is:
[0028] where α k represents the kth pixel in a circle with pixel blur radius r, Num represents the total number of pixels in the corresponding circle, β represents the pixel value of each pixel after averaging the pixel values, r=1, 2, 3.
[0029] Furthermore, the process of establishing the safety early warning network includes:
[0030] The same number of event process nodes and warning nodes are set according to the number of process feature images and warning feature images corresponding to each event process feature video and warning feature video;
[0031] Input each event process feature image and warning feature image into the event process node and the warning node in sequence according to the time sequence of the segmentation, and connect the event process node and the warning node corresponding to the same campus dangerous event in sequence on the event time axis, thereby obtaining the event feature chain corresponding to the campus dangerous event;
[0032] Compare the event process nodes and warning nodes of the event feature chain of campus dangerous events in the same period, and set a continuous similarity quantity threshold;
[0033] If there are similar continuous event process nodes or continuous warning nodes in two event feature chains, and the number of nodes in the continuous event process nodes or continuous warning nodes is greater than or equal to the continuous similarity number threshold, the corresponding continuous event process nodes or continuous warning nodes on the two event feature chains are set with the same label, and the corresponding continuous event process nodes or continuous warning nodes are respectively integrated into associated nodes and connected;
[0034] Otherwise, no operation is performed, and the above process of generating and connecting related nodes in each event feature chain is repeated to obtain a safety warning network, and each event feature chain in the safety warning network is labeled with the corresponding event name, and the safety warning network is labeled with the time period.
[0035] Furthermore, the process of comparing the event process nodes and warning nodes of different event feature chains includes:
[0036] A two-dimensional coordinate system is established, and the event process nodes or feature processes in the warning nodes from different event feature chains are sequentially overlapped and mapped on the two-dimensional coordinate system;
[0037] The horizontal angle between pixels at adjacent positions on each feature process is obtained, and then the horizontal angle between pixels on different feature processes but in overlapping states is compared, and the overall contrast value is counted and the contrast threshold is set. The calculation process of the overall contrast value is: Where η represents the horizontal angle comparison value between pixels in different feature processes but in overlapping states, λ q and ω q They represent the horizontal angle between the qth pixels in the two feature processes, and P represents the total number of pixels in the two feature processes;
[0038] If the overall comparison value is greater than or equal to the comparison threshold, the event process nodes or warning nodes in the corresponding two event feature chains are judged to be similar, otherwise they are judged to be dissimilar.
[0039] Furthermore, the establishment process of the warning event decision includes:
[0040] Match the corresponding security warning network according to the collection period of the campus scene data, and generate several optical image data and infrared image data respectively according to the optical video data and infrared video data in the campus scene data;
[0041] The characteristic process of generating the event process characteristic image and the warning characteristic image is adopted to obtain the characteristic process in each optical image data and the infrared image data, and the characteristic processes in the optical image data and the infrared image data corresponding to the same time node and the same position are superimposed on each other, so as to obtain the infrared characteristic process of the corresponding position at each time node;
[0042] By using a process similar to comparing event process nodes and warning nodes of different event feature chains, the infrared feature process at each time node is matched with the event feature chain in the matching security warning network, and a matching quantity threshold is set;
[0043] If there is a continuous infrared feature process that matches the threshold number of continuous event process nodes or continuous warning nodes in any event feature chain, it is determined that the event corresponding to the event feature chain is expected to occur or is occurring in the corresponding group of optical cameras and infrared cameras, and the corresponding safety warning information is generated, otherwise no operation is performed;
[0044] When the security warning module generates security warning information, it locates the location of the event according to the numbers of the optical camera and the infrared camera in the security warning information, and then generates a warning event decision and sends it to the campus management personnel. The campus management personnel then handles the corresponding event according to the warning event decision.
[0045] Compared with the prior art, the present invention has the following beneficial effects:
[0046] The present invention generates event process characteristic images and warning characteristic images through a number of campus dangerous events, and extracts corresponding characteristic processes from the event process characteristic images and the warning characteristic images, and then establishes an event characteristic chain according to the characteristic process, establishes a safety warning network according to the correlation between different event characteristic chains, and marks the corresponding time periods, generates corresponding infrared characteristic processes according to campus scene data in each time period, inputs the infrared characteristic processes into the safety warning network, and then outputs corresponding safety warning information and generates corresponding warning event decisions, thereby achieving the improvement of the accuracy and efficiency of campus safety warnings while ensuring the accuracy of obtaining campus status data. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention.
[0048] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION
[0049] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.
[0050] like Figure 1 As shown, the smart campus safety warning system based on video twin behavior analysis includes a cloud computing platform, which is communicatively connected to a campus scene data acquisition module, a warning event analysis module and a safety warning module.
[0051] The campus scene data acquisition module is used to set class time periods, break time periods and holiday time periods, and install several cameras with overlapping shooting ranges on campus, so as to collect campus scene data sets in each time period through the cameras.
[0052] The warning event analysis module is used to generate event process feature images and warning feature images based on several pre-stored campus dangerous events, and extract the corresponding feature processes from the event process feature images and warning feature images, and then establish an event feature chain based on the feature process, establish a safety warning network based on the correlation between different event feature chains, and mark the corresponding time periods.
[0053] The safety warning module is used to generate corresponding infrared feature processes according to campus scene data in various time periods, input the infrared feature processes into the safety warning network, and then output corresponding safety warning information and generate corresponding warning event decisions.
[0054] Further, the working principle of the present invention is described below by way of examples:
[0055] The staff uploads the semester table to the campus scene data collection module, and the semester table records the daily course names, break times, school dismissal times and holiday times during the class period in the corresponding semester.
[0056] Then, the campus scene data collection module establishes a semester timeline according to the length of the semester schedule, and divides the semester timeline into several class periods, break periods, and holiday periods.
[0057] At the same time, the campus scene data acquisition module installs several groups of optical cameras and infrared cameras at various locations on the campus, and sets the same number s1, s2, s3, ..., s n , where n is a natural number greater than 0.
[0058] It should be noted that the shooting range and shooting angle of each group of optical cameras and infrared cameras remain consistent at any time, and the shooting angle can be moved within a certain range. At the same time, the shooting range and shooting angle of each group of optical cameras and infrared cameras at adjacent spatial positions have an intersection relationship at any time.
[0059] A data upload cycle is set, wherein the duration of the data upload cycle is less than half of the duration of the break period, and each time a data upload cycle begins, each group of optical cameras and infrared cameras captures the optical video data and infrared video data within its shooting range, and sends the optical video data and infrared video data to the campus data acquisition module at the end of the data upload cycle.
[0060] Whenever a data upload cycle ends, the campus data acquisition module marks each group of optical video data and infrared video data with the number of the corresponding optical camera and infrared camera, and integrates all optical video data and infrared video data to generate the corresponding campus scene data set, and marks the corresponding acquisition time.
[0061] Then the campus scene data collection module sends the campus scene data set to the safety warning module.
[0062] Furthermore, the early warning event analysis module is preset with several campus dangerous events, and the campus dangerous events include event name, time period, event process feature video and warning feature video.
[0063] The event names include fire, campus violence, natural disasters, etc.
[0064] The characteristic video of the event process includes excessively high temperature in the area, physical conflict between people, and precipitation in the area.
[0065] The warning feature video includes abnormal and severe temperature rise in the area, abnormal smoke, etc.
[0066] An event timeline is established and several time nodes are set on the event timeline. At the same time, the early warning event analysis module establishes the event process characteristics and warning characteristics according to each campus dangerous event to establish a corresponding event feature chain.
[0067] The process of establishing the event feature chain includes:
[0068] The early warning event analysis module divides the event process feature video and the warning feature video into a number of event process feature images and warning feature images by frame.
[0069] The event process characteristic image and the warning characteristic image are grayed out respectively, and the pixel value of each pixel in the event process characteristic image and the warning characteristic image is obtained. The pixel blur radius r is set, and each pixel in the event process characteristic image and the warning characteristic image is selected as the circle center pixel respectively. Then, the pixel value of the circle center pixel and all pixels within the selected pixel blur radius r are averaged, where the pixel value averaging formula is: where α k represents the kth pixel in a circle with pixel blur radius r, Num represents the total number of pixels in the corresponding circle, β represents the pixel value of each pixel after averaging the pixel values, r=1, 2, 3.
[0070] It should be noted that if the pixels around the circle center pixel cannot form a circle with a pixel blur radius r, the next pixel is selected as the circle center pixel until all pixels in the event process feature image and the warning feature image are traversed.
[0071] A characteristic pixel value threshold is set, and then the pixel value of each pixel in the event process characteristic image and the warning characteristic image after averaging each pixel is compared with the characteristic pixel value threshold.
[0072] Pixels with values greater than or equal to the characteristic pixel threshold are recorded as characteristic pixels, and no operation is performed on the remaining pixels.
[0073] The feature pixels at adjacent positions are connected in sequence to obtain the feature processes of the feature images of each event process and the feature process of the warning feature images.
[0074] The same number of event process nodes and warning nodes are set according to the number of process feature images and warning feature images corresponding to each event process feature video and warning feature video;
[0075] Each event process feature image and warning feature image is input into the event process node and warning node in sequence according to the time sequence of segmentation, and the event process nodes and warning nodes corresponding to the same campus dangerous event are connected in sequence on the event timeline in time sequence, so as to obtain the event feature chain corresponding to the campus dangerous event.
[0076] The event process nodes and warning nodes of the event feature chain of campus dangerous events in the same period are compared, and a continuous similarity quantity threshold is set.
[0077] It should be noted that the process of comparing event process nodes and warning nodes of different event feature chains includes:
[0078] A two-dimensional coordinate system is established, and the event process nodes from different event feature chains or the feature processes in the warning nodes are sequentially overlapped and mapped on the two-dimensional coordinate system.
[0079] The horizontal angle between pixels at adjacent positions on each feature process is obtained, and then the horizontal angle between pixels on different feature processes but in overlapping states is compared, and the overall contrast value is counted and the contrast threshold is set. The calculation process of the overall contrast value is: Where η represents the horizontal angle comparison value between pixels in different feature processes but in overlapping states, λ q and ω q They respectively represent the horizontal angle between the qth pixels in the two feature processes, and P represents the total number of pixels in the two feature processes.
[0080] If the overall comparison value is greater than or equal to the comparison threshold, the event process nodes or warning nodes in the corresponding two event feature chains are judged to be similar, otherwise they are judged to be dissimilar.
[0081] If there are similar continuous event process nodes or continuous warning nodes in two event feature chains, and the number of nodes in the continuous event process nodes or continuous warning nodes is greater than or equal to the continuous similarity number threshold, then the corresponding continuous event process nodes or continuous warning nodes on the two event feature chains are set with the same label, and the corresponding continuous event process nodes or continuous warning nodes are respectively integrated into associated nodes and connected.
[0082] Otherwise, no operation is performed, and the above process of generating and connecting related nodes in each event feature chain is repeated to obtain a safety warning network, and each event feature chain in the safety warning network is labeled with the corresponding event name, and the safety warning network is labeled with the time period.
[0083] Furthermore, the warning event analysis module sends the security warning network to the security warning module, and then the security warning module matches the corresponding security warning network according to the collection period of the campus scene data.
[0084] At the same time, several optical image data and infrared image data are generated respectively according to the optical video data and infrared video data in the campus scene data.
[0085] The characteristic process of generating event process characteristic images and warning characteristic images is adopted to obtain the characteristic processes in each optical image data and infrared image data, and the characteristic processes in the optical image data and infrared image data corresponding to the same time node and the same position are superimposed on each other, so as to obtain the infrared characteristic process of the corresponding position at each time node.
[0086] By using a process similar to comparing event process nodes and warning nodes of different event feature chains, the infrared feature process at each time node is matched with the event feature chain in the matching security warning network, and a matching quantity threshold is set.
[0087] If there is a continuous infrared feature process that matches the threshold number of continuous event process nodes or continuous warning nodes in any event feature chain, the corresponding group of optical cameras and infrared cameras is judged to be expected to occur or the event corresponding to the event feature chain is occurring, and the corresponding safety warning information is generated, otherwise no action is taken.
[0088] The safety warning information includes the event name of the event that is expected to occur or is occurring and the numbers of the corresponding groups of optical cameras and infrared cameras.
[0089] Furthermore, when the security warning module generates security warning information, it locates the location of the event according to the numbers of the optical camera and the infrared camera in the security warning information, and then generates a warning event decision and sends it to the campus management personnel, who then handle the corresponding event according to the warning event decision.
[0090] The warning event decision includes the event location and event name.
[0091] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A smart campus safety early warning system based on video twin behavior analysis, including a cloud computing platform, characterized in that: The cloud computing platform is communicatively connected with a campus scene data acquisition module, a warning event analysis module and a safety warning module; The campus scene data collection module is used to set the class time period, break time period and holiday time period, and install a number of cameras with overlapping shooting ranges on the campus, so as to collect the campus scene data set in each time period through the cameras; The warning event analysis module is used to generate event process feature images and warning feature images based on a number of pre-stored campus dangerous events, and extract the corresponding feature process from the event process feature images and the warning feature images, and then establish an event feature chain based on the feature process, establish a safety warning network based on the correlation between different event feature chains, and mark the corresponding time period; The safety warning module is used to generate corresponding infrared feature processes according to campus scene data in various time periods, input the infrared feature processes into the safety warning network, and then output corresponding safety warning information and generate corresponding warning event decisions.
2. The smart campus safety early warning system based on video twin behavior analysis according to claim 1 is characterized in that: The collection process of the campus scene data set includes: The staff uploads the semester table to the campus scene data collection module, and the semester table records the daily course names, break times, dismissal times, and vacation times during the corresponding semester; Then the campus scene data collection module establishes a semester timeline according to the length of the semester table, and divides the semester timeline into several class periods, break periods, and holiday periods; At the same time, the campus scene data acquisition module installs several groups of optical cameras and infrared cameras at various locations on the campus, and sets the same number for the optical cameras and infrared cameras in the same group; A data upload cycle is set, wherein the duration of the data upload cycle is less than half of the duration of the break period, and each group of optical cameras and infrared cameras captures optical video data and infrared video data within their shooting ranges at the beginning of each data upload cycle, and sends the optical video data and infrared video data to the campus data acquisition module at the end of the data upload cycle; Whenever a data upload cycle ends, the campus data acquisition module marks each set of optical video data and infrared video data with the number of the corresponding optical camera and infrared camera, and integrates all the optical video data and infrared video data to generate the corresponding campus scene data set.
3. The smart campus safety early warning system based on video twin behavior analysis according to claim 2 is characterized in that: The shooting range and shooting angle of each group of optical cameras and infrared cameras remain consistent at any time, and at the same time, the shooting range and shooting angle of each group of optical cameras and infrared cameras at adjacent spatial positions have an intersecting relationship at any time.
4. The smart campus safety early warning system based on video twin behavior analysis according to claim 3 is characterized in that: The feature extraction process includes: The early warning event analysis module is preset with several campus dangerous events, and the campus dangerous events include event names, time periods, event process feature videos, and warning feature videos; An event timeline is established, and several time nodes are set on the event timeline. The early warning event analysis module divides the event process feature video and the warning feature video into several event process feature images and warning feature images by frame; The event process characteristic image and the warning characteristic image are grayed out respectively, and the pixel value of each pixel in the event process characteristic image and the warning characteristic image is obtained, and the pixel blur radius r is set, and each pixel in the event process characteristic image and the warning characteristic image is selected as the circle center pixel respectively, and then the pixel value of the circle center pixel and all pixels within the selected pixel blur radius r are averaged; Setting a characteristic pixel value threshold, and then comparing the pixel value of each pixel in the event process characteristic image and the warning characteristic image after averaging each pixel with the characteristic pixel value threshold; Pixels with values greater than or equal to the characteristic pixel threshold are recorded as characteristic pixels, and no operation is performed on the remaining pixels; The feature pixels at adjacent positions are connected in sequence to obtain the feature processes of the feature images of each event process and the feature process of the warning feature images.
5. The smart campus safety early warning system based on video twin behavior analysis according to claim 4 is characterized in that: The pixel value averaging formula is: where α k represents the kth pixel in a circle with pixel blur radius r, Num represents the total number of pixels in the corresponding circle, β represents the pixel value of each pixel after averaging the pixel values, r=1, 2, 3.
6. The smart campus safety early warning system based on video twin behavior analysis according to claim 4 is characterized in that: The process of establishing the safety early warning network includes: The same number of event process nodes and warning nodes are set according to the number of process feature images and warning feature images corresponding to each event process feature video and warning feature video; Input each event process feature image and warning feature image into the event process node and the warning node in sequence according to the time sequence of the segmentation, and connect the event process node and the warning node corresponding to the same campus dangerous event in sequence on the event time axis, thereby obtaining the event feature chain corresponding to the campus dangerous event; Compare the event process nodes and warning nodes of the event feature chain of campus dangerous events in the same period, and set a continuous similarity quantity threshold; If there are similar continuous event process nodes or continuous warning nodes in two event feature chains, and the number of nodes in the continuous event process nodes or continuous warning nodes is greater than or equal to the continuous similarity number threshold, the corresponding continuous event process nodes or continuous warning nodes on the two event feature chains are set with the same label, and the corresponding continuous event process nodes or continuous warning nodes are respectively integrated into associated nodes and connected; Otherwise, no operation is performed, and the above process of generating and connecting related nodes in each event feature chain is repeated to obtain a safety warning network, and each event feature chain in the safety warning network is labeled with the corresponding event name, and the safety warning network is labeled with the time period.
7. The smart campus safety early warning system based on video twin behavior analysis according to claim 6 is characterized in that: The process of comparing event process nodes and warning nodes of different event feature chains includes: A two-dimensional coordinate system is established, and the event process nodes or feature processes in the warning nodes from different event feature chains are sequentially overlapped and mapped on the two-dimensional coordinate system; The horizontal angle between pixels at adjacent positions on each feature process is obtained, and then the horizontal angle between pixels on different feature processes but in overlapping states is compared, and the overall contrast value is counted and the contrast threshold is set. The calculation process of the overall contrast value is: Where η represents the horizontal angle comparison value between pixels in different feature processes but in overlapping states, λ q and ω q They represent the horizontal angle between the qth pixels in the two feature processes, and P represents the total number of pixels in the two feature processes; If the overall comparison value is greater than or equal to the comparison threshold, the event process nodes or warning nodes in the corresponding two event feature chains are judged to be similar, otherwise they are judged to be dissimilar.
8. The smart campus safety early warning system based on video twin behavior analysis according to claim 6 is characterized in that: The establishment process of the warning event decision includes: Match the corresponding security warning network according to the collection period of the campus scene data, and generate several optical image data and infrared image data respectively according to the optical video data and infrared video data in the campus scene data; The characteristic process of generating the event process characteristic image and the warning characteristic image is adopted to obtain the characteristic process in each optical image data and the infrared image data, and the characteristic processes in the optical image data and the infrared image data corresponding to the same time node and the same position are superimposed on each other, so as to obtain the infrared characteristic process of the corresponding position at each time node; By using a process similar to comparing event process nodes and warning nodes of different event feature chains, the infrared feature process at each time node is matched with the event feature chain in the matching security warning network, and a matching quantity threshold is set; If there is a continuous infrared feature process that matches the threshold number of continuous event process nodes or continuous warning nodes in any event feature chain, it is determined that the event corresponding to the event feature chain is expected to occur or is occurring in the corresponding group of optical cameras and infrared cameras, and the corresponding safety warning information is generated, otherwise no operation is performed; When the security warning module generates security warning information, it locates the location of the event according to the numbers of the optical camera and the infrared camera in the security warning information, and then generates a warning event decision and sends it to the campus management personnel. The campus management personnel then handles the corresponding event according to the warning event decision.
