Video security event recognition and early warning system based on artificial intelligence

Abstract

The present invention belongs to the technical field of video security event recognition, and specifically relates to a video security event recognition and early warning system based on artificial intelligence. It performs panoramic image synthesis processing on the video data collected by each controlled camera in the target construction area, obtains the panoramic monitoring video stream of the target construction area in real time, accurately identifies the type of construction machinery currently in the target construction area and tracks the position changes of dangerous feature points, analyzes its current collision risk coefficient, rollover risk coefficient and personnel violation risk coefficient to determine whether there is a security incident at present. If it is determined to be so, it further identifies the current security incident type, combines the risk impact of the construction machinery type and the risk impact of the security incident type relative to the current construction stage, and assesses the current security incident risk level to carry out early warning work, effectively providing continuous, immediate and accurate safety monitoring, thereby greatly improving the level and efficiency of video security incident management in the construction area.

Description

Technical Field

[0001] The present invention belongs to the technical field of video security event recognition, and specifically relates to a video security event recognition and early warning system based on artificial intelligence. Background Art

[0002] As construction sites expand in scale and become increasingly complex, traditional safety incident management relies on manual inspections and post-event intervention. This lack of response time and limited monitoring makes it difficult to adapt to the complex demands of modern construction. To effectively mitigate the significant risks these incidents pose to construction workers, project progress, and economic profitability, the introduction of intelligent machine vision and artificial intelligence technologies for rapid identification and early warning of safety incidents at construction sites is crucial.

[0003] There are also some solutions related to video security event monitoring and identification at construction sites in the existing technology. For example, the construction site construction event monitoring system based on video surveillance identification, published in China with patent number CN117640892A, includes a data acquisition system, a management system, and a management terminal. The data acquisition system collects live video of the construction site in real time and transmits the live video and location information to the management system. The management system determines the construction site map based on the location information and transmits it to the management terminal. The live video is analyzed to determine whether a dangerous event has occurred at the construction site and the target location of the dangerous event. If a dangerous event occurs, a corresponding alarm message is generated and transmitted to the data acquisition system and the management terminal. The data acquisition system controls the laser emitter to turn to the target location and emit light. The management terminal issues an alarm based on the alarm message. This solves the problem that existing construction site alarm systems do not clearly indicate the location of the dangerous event when issuing an alarm, resulting in managers being unable to handle it in a timely manner, thus endangering the safety of workers.

[0004] Another Chinese patent publication number CN116684555A is a method and system for collecting on-site comprehensive data at construction sites. It uses cameras in a video surveillance system to monitor the construction site and entrance and exit locations of the construction site, and can play the video on the terminal display screen, so that personnel can observe and control the display screen and perform early warning processing. The images captured by the camera can warn of whether there are dangerous situations at the construction site. Through visual display technology and comprehensive real-time data collection and analysis from all directions, it can achieve panoramic monitoring of the construction site, real-time monitoring and safety protection of key areas, and real-time monitoring and analysis of the behavior of site personnel, thereby preventing the occurrence of construction site accidents. It also adopts information technology to further strengthen the effective system management of the site, and uses intelligence to achieve the effects and advantages of cloud supervision.

[0005] Although the above scheme proposes some related solutions involving video security event monitoring and identification at construction sites, the existing technology still has the following limitations, specifically: 1. The existing technology for panoramic monitoring of construction sites mainly relies on all-round deployment of cameras to piece together the panorama. During video analysis, multiple directional images need to be processed one by one, resulting in a lengthy analysis process and high consumption of computing resources. In addition, multi-image analysis doubles the pressure on data transmission and storage, making it difficult to meet real-time requirements and not conducive to efficient and accurate identification of construction site safety incidents.

[0006] 2. The construction site environment is complex, and there are many types of machinery and equipment. When construction machinery causes a safety incident, it is very likely to trigger a chain reaction, which will not only cause casualties and property losses, but may also affect the progress and quality of the project. However, existing technologies have relatively ignored the consideration of safety incidents related to construction machinery on the construction site.

[0007] 3. Existing technologies fail to fully consider the differences in safety incident types at different construction stages, which may result in emergency response measures being unable to respond to safety incident types in a detailed and effective manner, thereby affecting the effectiveness of emergency response. Summary of the Invention

[0008] In order to overcome the shortcomings of the background technology, the embodiments of the present invention provide a video security event recognition and early warning system based on artificial intelligence, which can effectively solve the problems involved in the above background technology.

[0009] The purpose of the present invention can be achieved through the following technical solutions: a video security event recognition and early warning system based on artificial intelligence, including: a video synthesis and streaming module, a construction machinery recognition module, a security event judgment module, a security event early warning module and a cloud database.

[0010] The video synthesis and streaming module is connected to the construction machinery identification module, the construction machinery identification module is connected to the security event judgment module, the security event judgment module is connected to the security event early warning module, and the cloud database is connected to the construction machinery identification module, the security event judgment module, and the security event early warning module respectively.

[0011] The video synthesis and stream acquisition module is used to perform panoramic image synthesis processing on the video data collected by each control camera in the target construction area to obtain the panoramic monitoring video stream of the target construction area in real time.

[0012] The construction machinery identification module is used to identify the type of construction machinery in the current target construction area based on the real-time panoramic monitoring video stream of the target construction area, mark it as the target construction machinery, and capture the various dangerous feature points on the target construction machinery.

[0013] The safety incident judgment module is used to continuously track the position changes of various dangerous feature points on the target construction machinery and determine whether a safety incident has occurred on the current target construction machinery.

[0014] The safety incident warning module is used to identify the current safety incident type and assess the current safety incident risk level when it is determined that a safety incident has occurred in the current target construction machinery, so as to carry out early warning work.

[0015] The cloud database is used to store preset standard three-dimensional framework outlines and preset standard operation trajectories of various types of construction machinery, store the reference distances of various dangerous feature points marked in the area scope of preset standard three-dimensional framework outlines of various types of construction machinery and their relative center of gravity positions, store the preset risk impact weights corresponding to safety incidents caused by various types of construction machinery, store the preset risk impact weights corresponding to each type of safety incident for each construction stage, store the preset collision risk coefficient warning threshold, preset rollover risk coefficient warning threshold, preset personnel violation risk coefficient warning threshold, and store the preset benchmark risk level range corresponding to each risk level of construction safety incidents.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention obtains a panoramic monitoring video stream of the target construction area in real time by performing panoramic image synthesis processing on the video data collected by each camera deployed in the target construction area. Compared with the decentralized processing of each camera image, the present invention reduces the waste of computing resources while improving the monitoring operation efficiency and response accuracy, thereby enabling more timely detection of abnormal situations.

[0017] (2) The present invention accurately identifies the type of construction machinery currently located in the target construction area and tracks the position changes of dangerous feature points, analyzes its current collision risk coefficient, rollover risk coefficient and personnel violation risk coefficient to determine whether there is a safety incident at present, and effectively realizes the quantitative presentation of the safety risks of construction machinery at the construction site, helps to intuitively grasp the risk situation and formulate response strategies in a timely manner.

[0018] (3) The present invention identifies the current safety event type, combines the risk impact of the construction machinery type and the risk impact of the safety event type relative to the current construction stage, and assesses the current safety event risk level to carry out early warning work. It provides adaptive early warning measures for various safety events that occur with different machinery and at different construction stages on the construction site, greatly improving the reliability and timeliness of construction site safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0020] Figure 1 Schematic diagram of module connection of the present invention.

[0021] Figure 2 This is a logical diagram for the present invention to identify the type of construction machinery currently located in the target construction area.

[0022] Figure 3 This is a logical diagram of the present invention for determining whether a safety incident has occurred on the current target construction machinery. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] Reference Figure 1 As shown, the present invention provides a video security event recognition and early warning system based on artificial intelligence, including: a video synthesis and streaming module, a construction machinery recognition module, a security event judgment module, a security event early warning module and a cloud database.

[0025] The video synthesis and streaming module is connected to the construction machinery identification module, the construction machinery identification module is connected to the security event judgment module, the security event judgment module is connected to the security event early warning module, and the cloud database is connected to the construction machinery identification module, the security event judgment module, and the security event early warning module respectively.

[0026] The video synthesis and stream acquisition module is used to perform panoramic image synthesis processing on the video data collected by each control camera in the target construction area to obtain a panoramic monitoring video stream of the target construction area in real time.

[0027] Specifically, the real-time acquisition of the panoramic monitoring video stream of the target construction area includes: performing internal parameter calibration and external parameter calibration for each controlled camera in the target construction area, establishing a proportional relationship between the image plane coordinates of each controlled camera and the preset three-dimensional coordinates, and obtaining the relative position relationship between each controlled camera.

[0028] The video frame images captured by each controlled camera at the same time point are preprocessed, and the reference feature points in the video frame images captured by each controlled camera at the same time point are retrieved and their corresponding feature descriptors are calculated. Through the mutual matching operation of feature descriptors, the reference feature point pairs between the video frame images captured by each controlled camera at the same time point are retrieved. Combined with the relative position relationship between each controlled camera, the video frame images captured by each controlled camera at the same time point are subjected to image registration processing and image fusion processing, and the fused video frame images are output according to the preset video frame rate requirements and preset video encoding format requirements, so as to obtain the panoramic monitoring video stream of the target construction area in real time.

[0029] The embodiment of the present invention performs panoramic image synthesis processing on the video data collected by each controlled camera in the target construction area, and obtains a panoramic monitoring video stream of the target construction area in real time. Compared with the decentralized processing of each camera image, it reduces the waste of computing resources while improving the monitoring operation efficiency and response accuracy, so that abnormal situations can be discovered more promptly.

[0030] The construction machinery identification module is used to identify the type of construction machinery currently in the target construction area based on the real-time acquired panoramic monitoring video stream of the target construction area, mark it as the target construction machinery, and capture various dangerous feature points on the target construction machinery.

[0031] Reference Figure 2 As shown, specifically, the identification of the type of construction machinery in the current target construction area includes: extracting each video frame in the panoramic monitoring video stream of the current target construction area, identifying the construction machinery area in each video frame in the panoramic monitoring video stream of the current target construction area based on the color feature difference and texture feature difference of the construction machinery relative to the image background, and outlining the three-dimensional frame contour of the construction machinery based on three-dimensional reconstruction, and analyzing the similarity of the three-dimensional frame contour of the construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area with respect to the preset standard three-dimensional frame contours of various types of construction machinery stored in the cloud database. ,in For the serial numbers of various types of construction machinery, , is the number of each video frame in the panoramic monitoring video stream of the current target construction area, .

[0032] A preset number of pixel points on the three-dimensional frame outline of construction machinery are randomly selected from the initial video frame image in the panoramic monitoring video stream of the current target construction area and tracked and marked. According to the position changes of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area, the operation trajectory of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area is obtained, and the similarity of the operation trajectory of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area to the preset standard operation trajectory of various types of construction machinery stored in the cloud database is analyzed. , is the number of each pixel point on the 3D frame outline of the construction machinery, .

[0033] It should be noted that the above 、 Both involve shape matching and contour comparison, which can be imported into image software and directly output by computer vision tools.

[0034] Analyze the matching confidence between the construction machinery in the current target construction area and various types of construction machinery , select the construction machinery type corresponding to the maximum matching confidence as the construction machinery type of the current target construction area.

[0035] Specifically, the ,in The preset weights are respectively the similarity of the 3D frame outline of the construction machinery and the similarity of the operation trajectory. is the number of video frames in the panoramic monitoring video stream of the current target construction area, The operation trajectory of each pixel point on the 3D frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area relative to the first pixel point stored in the cloud database. The maximum value among the similarities of preset standard operation trajectories of similar construction machinery.

[0036] Specifically, the capturing of each dangerous feature point on the target construction machinery includes: extracting the reference distances of each dangerous feature point marked in the preset standard three-dimensional framework outline area scope of the target construction machinery stored in the cloud database and their relative center of gravity positions, aligning the three-dimensional framework outline of the target construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area with the preset standard three-dimensional framework outline of the target construction machinery type, obtaining the potential comparison pixel points of each dangerous feature point marked in the three-dimensional framework outline area scope of the target construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area relative to the preset standard three-dimensional framework outline area scope, and screening the best comparison pixel points of each dangerous feature point marked in the three-dimensional framework outline area scope of the target construction machinery in the video frame in the panoramic monitoring video stream of the current target construction area relative to the preset standard three-dimensional framework outline area scope according to the distance between the potential comparison pixel points and the center of gravity position of the area scope, so as to realize the capture of each dangerous feature point on the target construction machinery.

[0037] The safety event judgment module is used to continuously track the position changes of each dangerous feature point on the target construction machinery and judge whether a safety event occurs in the target construction machinery.

[0038] Reference Figure 3 As shown, specifically, the determination of whether a safety incident occurs in the current target construction machinery includes: identifying each fixed building facility in the panoramic monitoring video stream of the current target construction area, obtaining the acceleration fluctuation degree of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area, and the shortest distance relative to each fixed building facility. and velocity pointing deviation , is the number of each dangerous feature point of the target construction machinery, , is the number of each fixed building facility, , analyze the collision risk coefficient of the current target construction machinery.

[0039] It should be noted that the specific analysis process of the collision risk coefficient of the current target construction machinery is as follows: calculate the potential collision direction of each dangerous feature point of the target construction machinery relative to each fixed building facility in the panoramic monitoring video stream of the current target construction area , , The preset safety isolation distance threshold of the target construction machinery operation relative to the fixed building facilities is used to screen the maximum potential collision directionality of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area relative to the fixed building facilities, and multiply it with the acceleration fluctuation degree of the corresponding dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area to obtain the comprehensive collision directionality of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area relative to the fixed building facilities. The maximum comprehensive collision directionality of the dangerous feature points relative to the fixed building facilities is further screened as the collision risk coefficient of the current target construction machinery.

[0040] It should also be noted that the theoretical basis for analyzing the collision risk coefficient of the aforementioned target construction machinery is as follows: First, the changes in the acceleration of dangerous feature points in the video frame image are of great significance. At the moment of a collision, such as when a construction machine collides with a fixed structure, the mechanical components are subjected to a reaction force from the fixed structure at the moment of contact, causing their original state of motion to be drastically altered. Just before the collision, the target construction machinery continues to approach the fixed structure, with the distance continuously decreasing. However, at the moment of contact, the component speed changes dramatically in a very short period of time, causing the acceleration of the dangerous feature points to exhibit large pulse values—fluctuations that suddenly increase and then decrease rapidly. For example, when the construction machine's boom is swinging close to a nearby building, the acceleration of the dangerous feature point at the boom's endpoint will experience large fluctuations due to the impact force generated by the collision.

[0041] Secondly, safety clearance is a critical distance standard for avoiding collisions and ensuring construction safety. When the distance between a hazardous feature and a fixed structure falls below this safety distance, it indicates that the structure is within a very close danger zone. This is an intuitive and crucial prerequisite for determining the likelihood of a collision. Once within this danger zone, the probability of a collision increases significantly.

[0042] Third, the direction of the danger feature's velocity is also a key consideration. If the velocity of a danger feature is pointing toward a fixed structure, it indicates the target construction machinery is moving toward a potential collision, increasing the likelihood of a collision.

[0043] In summary, when the three conditions are met simultaneously: the distance between the dangerous feature point and the target is less than the safe distance, the speed direction points to the target, and the acceleration fluctuation is particularly large, it is highly likely that a collision event is occurring or is about to occur.

[0044] Obtain the degree of deviation of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area relative to the center of gravity of the three-dimensional frame contour area , the maximum change rate of the relative position angle between dangerous feature points , according to the formula Analyze the rollover risk coefficient of the current target construction machinery, where Indicates that the target construction machinery in the panoramic monitoring video stream of the current target construction area is The deviation degree of the dangerous feature point relative to the center of gravity of the three-dimensional framework contour area and the preset allowable deviation degree threshold of the target construction machinery relative to the center of gravity The calculated difference is compared with 0, and the larger value is taken.

[0045] It should be noted that the above The specific acquisition process is as follows: the spacing and angle of each dangerous feature point of the target construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area relative to the center of gravity position of the three-dimensional framework contour area, the absolute value of the spacing is subtracted from the reference spacing, and the absolute difference is further analyzed by ratio between the reference spacing to obtain the spacing offset degree. Similarly, the angle is compared with the reference angle to obtain the angle offset degree, and the accumulated value of the angle offset degree and the spacing offset degree is used as the comprehensive offset degree. The average value of the comprehensive offset degree of each dangerous feature point of the target construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area relative to the center of gravity position of the three-dimensional framework contour area is calculated to obtain the offset degree of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area relative to the center of gravity position of the three-dimensional framework contour area. The reference angle of each dangerous feature point of the target construction machinery relative to the center of gravity position of the three-dimensional framework contour area can be specifically referred to the technical specification manual provided by the target construction machinery manufacturer to specify the angle of each dangerous feature point relative to the center of gravity position of the three-dimensional framework contour area under normal operating conditions.

[0046] above The specific acquisition process is as follows: randomly select a dangerous feature point of the target construction machinery, and pair it with other dangerous feature points in turn to obtain each pair of dangerous feature points of the target construction machinery, obtain the relative position angle of each pair of dangerous feature points of the target construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area, compare the relative position angle of a certain dangerous feature point pair in each video frame, extract the maximum and minimum values ​​and make a difference, perform a ratio analysis on the calculated difference and the preset reference relative position angle change value to obtain the change rate of the dangerous feature point pair of the target construction machinery in the panoramic monitoring video stream of the current target construction area, and similarly obtain the change rate of each pair of dangerous feature points of the target construction machinery in the panoramic monitoring video stream of the current target construction area, and select the maximum value as the maximum change rate of the relative position angle between the dangerous feature points of the target construction machinery in the panoramic monitoring video stream of the current target construction area.

[0047] It should be noted in particular that the preset allowable deviation threshold of the above-mentioned target construction machinery relative to the center of gravity can be specifically referred to the average allowable safe deviation of the dangerous characteristic points relative to the center of gravity recorded in the technical specification manual of the target construction machinery after testing by the target construction machinery manufacturer during the design and production process.

[0048] It should also be noted that the theoretical basis for analyzing the tipping risk coefficient of the aforementioned current target construction machinery is that when the construction machinery is in a stable state, the vertical line of its center of gravity falls within the support surface. Once a tipping tendency occurs, such as when the boom is overextended or an overweight object is lifted, the dangerous feature point will shift relative to the center of gravity. From a distance perspective, when the distance offset causes the vertical line of the center of gravity to gradually approach the edge of the support surface, the machine is in an unstable critical state. From an angular perspective, the change in the angle between the dangerous feature point and the center of gravity reflects the degree of tilt of the machine. As the angle increases, it means that the machine's tilt is increasing. When the angle reaches a certain level, exceeding the maximum angle at which the machine can maintain balance, tipping will occur.

[0049] Identify the construction workers in the panoramic surveillance video stream of the current target construction area, and obtain the shortest distance between the dangerous feature points of the target construction machinery and the construction workers in the panoramic surveillance video stream of the current target construction area , The number of each construction worker, , analyze the personnel violation risk coefficient of the current target construction machinery.

[0050] The collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery are compared with the preset collision risk coefficient warning threshold, preset rollover risk coefficient warning threshold, and preset personnel violation risk coefficient warning threshold stored in the cloud database to determine whether a safety incident has occurred with the current target construction machinery.

[0051] Specifically, the specific analysis formula of the personnel violation risk coefficient of the current target construction machinery is: ,in The relative position of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area The minimum value among the shortest distances between construction workers, It is the preset safety isolation distance threshold between the target construction machinery and the construction personnel.

[0052] It should be noted in particular that the preset safety isolation distance thresholds for the above-mentioned target construction machinery operations relative to fixed construction facilities and construction personnel can be obtained by referring to the minimum safety isolation distances for various types of construction machinery operating at the construction site relative to fixed construction facilities and construction personnel as stipulated in the Safety Technical Regulations for the Use of Construction Machinery, and matching the target construction machinery types.

[0053] Specifically, the judgment condition for the occurrence of a safety incident of the current target construction machinery is that any one of the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery is greater than or equal to its corresponding preset warning threshold.

[0054] The embodiment of the present invention accurately identifies the type of construction machinery currently in the target construction area and tracks the position changes of dangerous feature points, analyzes its current collision risk coefficient, rollover risk coefficient and personnel violation risk coefficient to determine whether there is a safety incident at present, and effectively realizes the quantitative presentation of the safety risks of construction machinery at the construction site, helps to intuitively grasp the risk situation and formulate response strategies in a timely manner.

[0055] The safety event warning module is used to identify the current safety event type and assess the current safety event risk level when it is determined that a safety event has occurred in the current target construction machinery, so as to carry out early warning work.

[0056] Specifically, the identification of the current safety event type includes: using the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the target construction machinery as quantitative indicators of the possibility of collision events, rollover events, and personnel violation approach events, respectively.

[0057] Filter the coefficient indicator objects among the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery that are greater than or equal to their corresponding preset warning thresholds, mark them as abnormal coefficient indicators, and trace the event type corresponding to the abnormal coefficient indicator to the current safety event type.

[0058] Specifically, the assessment of the current security event risk level includes: subtracting the abnormal coefficient index of the current security event type from its corresponding preset warning threshold, and taking the ratio of the calculated difference to its corresponding preset warning threshold as the current security event risk ratio. .

[0059] Extract the preset risk impact weights corresponding to the safety incidents caused by the target construction machinery type stored in the cloud database And the preset risk impact weight of the current construction stage of the target construction area where the current safety event type is located , according to the formula Analyze the baseline risk level of the current safety incident and assess the current safety incident risk level according to the preset baseline risk level range corresponding to each risk level of construction safety incidents stored in the cloud database.

[0060] The embodiment of the present invention identifies the current security event type, combines the risk impact of the construction machinery type and the risk impact of the security event type relative to the current construction stage, assesses the current security event risk level to carry out early warning work, and provides adaptive early warning measures for various security events that occur with different machinery and at different construction stages on the construction site, greatly improving the reliability and timeliness of construction site safety management.

[0061] The cloud database is used to store preset standard three-dimensional framework outlines and preset standard operation trajectories of various types of construction machinery, store the reference distances of various dangerous feature points marked in the area scope of the preset standard three-dimensional framework outlines of various types of construction machinery and their relative center of gravity positions, store the preset risk impact weights corresponding to safety incidents caused by various types of construction machinery, store the preset risk impact weights corresponding to each safety incident type for each construction stage, store the preset collision risk coefficient warning threshold, the preset rollover risk coefficient warning threshold, the preset personnel violation risk coefficient warning threshold, and store the preset benchmark risk degree range corresponding to each risk level of construction safety incidents.

[0062] The data sources in the cloud database of this embodiment are shown in Table 1 below.

[0063] Table 1 Detailed description of data sources in cloud database

[0064]

[0065] The above content is merely an example and explanation of the structure of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the scope of protection of the present invention.

Claims

1. Artificial intelligence-based video security event recognition and early warning system, characterized by: include: The video synthesis and streaming module is used to perform panoramic image synthesis processing on the video data collected by each control camera in the target construction area to obtain the panoramic monitoring video stream of the target construction area in real time; The construction machinery identification module is used to identify the type of construction machinery currently in the target construction area based on the real-time panoramic monitoring video stream of the target construction area, mark it as the target construction machinery, and capture the various dangerous feature points on the target construction machinery; The safety event judgment module is used to continuously track the position changes of each dangerous feature point on the target construction machinery and determine whether a safety event has occurred on the current target construction machinery; The safety incident warning module is used to identify the current safety incident type and assess the current safety incident risk level when it is determined that a safety incident has occurred on the current target construction machinery, thereby carrying out early warning work; A cloud database is used to store preset standard three-dimensional framework outlines and preset standard operation trajectories of various types of construction machinery, store the reference distances of various dangerous feature points marked in the area scope of the preset standard three-dimensional framework outlines of various types of construction machinery and the center of gravity of the area scope, store the preset risk impact weights corresponding to safety incidents caused by various types of construction machinery, store the preset risk impact weights corresponding to each type of safety incident for each construction stage, store the preset collision risk factor warning thresholds, preset rollover risk factor warning thresholds, preset personnel violation risk factor warning thresholds, and store the preset benchmark risk level ranges corresponding to each risk level of construction safety incidents; Analyze the collision risk coefficient of the target construction machinery based on the acceleration fluctuation degree of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area, the shortest distance relative to each fixed building facility, and the speed direction deviation; Analyze the rollover risk coefficient of the target construction machinery based on the degree of deviation of each dangerous feature point of the target construction machinery from the center of gravity of the three-dimensional framework outline area in the panoramic monitoring video stream of the current target construction area, and the maximum change rate of the relative position angle between the dangerous feature points; Analyze the risk factor of personnel violations on the current target construction machinery based on the shortest distance between each dangerous feature point of the target construction machinery and each construction worker in the panoramic monitoring video stream of the current target construction area; Determine whether a safety incident has occurred with the current target construction machinery based on its collision risk factor, rollover risk factor, and personnel violation risk factor; The assessment of the current security incident risk level includes: subtracting the abnormal coefficient index of the current security incident type from its corresponding preset warning threshold, and taking the ratio of the calculated difference to its corresponding preset warning threshold as the current security incident risk ratio. ; Extract the preset risk impact weights corresponding to the safety incidents caused by the target construction machinery type stored in the cloud database And the preset risk impact weight of the current construction stage of the target construction area where the current safety event type is located , according to the formula Analyze the baseline risk level of the current safety incident and assess the current safety incident risk level according to the preset baseline risk level range corresponding to each risk level of construction safety incidents stored in the cloud database.

2. The artificial intelligence-based video security event recognition and early warning system according to claim 1 is characterized by: The real-time acquisition of the panoramic monitoring video stream of the target construction area includes: performing internal parameter calibration and external parameter calibration for each control camera in the target construction area, establishing a proportional relationship between the image plane coordinates of each control camera and the preset three-dimensional coordinates, and obtaining the relative position relationship between each control camera; The video frame images captured by each controlled camera at the same time point are preprocessed, and the reference feature points in the video frame images captured by each controlled camera at the same time point are retrieved and their corresponding feature descriptors are calculated. Through the mutual matching operation of feature descriptors, the reference feature point pairs between the video frame images captured by each controlled camera at the same time point are retrieved. Combined with the relative position relationship between each controlled camera, the video frame images captured by each controlled camera at the same time point are subjected to image registration processing and image fusion processing, and the fused video frame images are output according to the preset video frame rate requirements and preset video encoding format requirements, so as to obtain the panoramic monitoring video stream of the target construction area in real time.

3. The artificial intelligence-based video security event recognition and early warning system according to claim 1 is characterized by: The method for identifying the type of construction machinery in the current target construction area includes: extracting each video frame in the panoramic monitoring video stream of the current target construction area, identifying the construction machinery area in each video frame in the panoramic monitoring video stream of the current target construction area based on the color feature difference and texture feature difference of the construction machinery relative to the image background, and outlining the three-dimensional frame outline of the construction machinery based on three-dimensional reconstruction, and analyzing the similarity of the three-dimensional frame outline of the construction machinery in each video frame in the panoramic monitoring video stream of the current target construction area with the preset standard three-dimensional frame outlines of various types of construction machinery stored in the cloud database. ,in For the serial numbers of various types of construction machinery, , is the number of each video frame in the panoramic monitoring video stream of the current target construction area, ; A preset number of pixel points on the three-dimensional frame outline of construction machinery are randomly selected from the initial video frame image in the panoramic monitoring video stream of the current target construction area and tracked and marked. According to the position changes of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area, the operation trajectory of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area is obtained, and the similarity of the operation trajectory of each pixel point on the three-dimensional frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area to the preset standard operation trajectory of various types of construction machinery stored in the cloud database is analyzed. , is the number of each pixel point on the 3D frame outline of the construction machinery, ; Analyze the matching confidence between the construction machinery in the current target construction area and various types of construction machinery , select the construction machinery type corresponding to the maximum matching confidence as the construction machinery type of the current target construction area.

4. The artificial intelligence-based video security event recognition and early warning system according to claim 3 is characterized by: described ,in The preset weights are respectively the similarity of the 3D frame outline of the construction machinery and the similarity of the operation trajectory. is the number of video frames in the panoramic monitoring video stream of the current target construction area, The operation trajectory of each pixel point on the 3D frame outline of the construction machinery in the panoramic monitoring video stream of the current target construction area relative to the first pixel point stored in the cloud database. The maximum value among the similarities of preset standard operation trajectories of similar construction machinery.

5. The artificial intelligence-based video security event recognition and early warning system according to claim 3 is characterized by: The capturing of each dangerous feature point on the target construction machinery includes: extracting the reference distances of each dangerous feature point marked in the preset standard three-dimensional framework outline area scope of the target construction machinery stored in the cloud database and their relative center of gravity positions, aligning the three-dimensional framework outline of the target construction machinery in each video frame in the current target construction area panoramic monitoring video stream with the preset standard three-dimensional framework outline of the target construction machinery type, obtaining the potential comparison pixel points of each dangerous feature point marked in the three-dimensional framework outline area scope of the target construction machinery in each video frame in the current target construction area panoramic monitoring video stream relative to the preset standard three-dimensional framework outline area scope, and screening the best comparison pixel points of each dangerous feature point marked in the three-dimensional framework outline area scope of the target construction machinery in the video frame in the current target construction area panoramic monitoring video stream relative to the preset standard three-dimensional framework outline area scope according to the distance between the potential comparison pixel points and the center of gravity position of the area scope, so as to realize the capturing of each dangerous feature point on the target construction machinery.

6. The artificial intelligence-based video security event recognition and early warning system according to claim 5, characterized in that: Determining whether a safety incident has occurred with the current target construction machine also includes: identifying each fixed building facility in the panoramic surveillance video stream of the current target construction area, obtaining the acceleration fluctuation degree, the shortest distance relative to each fixed building facility, and the speed direction deviation of each dangerous feature point of the target construction machine in the panoramic surveillance video stream of the current target construction area, and analyzing the collision risk coefficient of the current target construction machine; Obtain the degree of deviation of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area relative to the center of gravity of the three-dimensional framework outline area, and the maximum change rate of the relative position angle between the dangerous feature points, and analyze the rollover risk coefficient of the current target construction machinery; Identify the construction workers in the panoramic surveillance video stream of the current target construction area, and obtain the shortest distance between the dangerous feature points of the target construction machinery and the construction workers in the panoramic surveillance video stream of the current target construction area , is the number of each dangerous feature point of the target construction machinery, , The number of each construction worker, ,Analyze the risk factor of personnel violations of the current target construction machinery; The collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery are compared with the preset collision risk coefficient warning threshold, preset rollover risk coefficient warning threshold, and preset personnel violation risk coefficient warning threshold stored in the cloud database to determine whether a safety incident has occurred with the current target construction machinery.

7. The artificial intelligence-based video security event recognition and early warning system according to claim 6, characterized in that: The specific analysis formula of the personnel violation risk coefficient of the current target construction machinery is: ,in The relative position of each dangerous feature point of the target construction machinery in the panoramic monitoring video stream of the current target construction area The minimum value among the shortest distances between construction workers, It is the preset safety isolation distance threshold between the target construction machinery and the construction personnel.

8. The artificial intelligence-based video security event recognition and early warning system according to claim 6, characterized in that: The judgment condition for the occurrence of a safety incident of the current target construction machinery is that any one of the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery is greater than or equal to its corresponding preset warning threshold.

9. The artificial intelligence-based video security event recognition and early warning system according to claim 6, characterized in that: The identifying of the current safety event type includes: using the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the target construction machinery as quantitative indicators of the likelihood of a collision event, a rollover event, and a personnel violation approach event respectively; Filter the coefficient indicator objects among the collision risk coefficient, rollover risk coefficient, and personnel violation risk coefficient of the current target construction machinery that are greater than or equal to their corresponding preset warning thresholds, mark them as abnormal coefficient indicators, and trace the event type corresponding to the abnormal coefficient indicator to the current safety event type.

Images