Space-time intelligent collaborative field operation safety risk prevention and control capability improvement and application research and development

Abstract

The invention relates to the technical field of AI image recognition, in particular to space-time intelligent collaborative field operation safety risk prevention and control capability improvement and application research and development. The method comprises the following steps: configuring relevant information of an actually used camera according to field requirements; providing a target detection model designed according to supervision requirements according to business requirements; and according to a detection requirement, setting an independent focusing target name for each path of camera so as to carry out real-time tracking combining algorithm service and ontology service. The invention aims to provide the time-space intelligent cooperative field operation safety risk prevention and control capability improvement and application research and development, the target object in the video stream is identified and tracked in combination with time and space information so as to facilitate real-time analysis, the condition of false detection of the target object can be greatly relieved in combination with a multi-frame identification result, and the detection efficiency is improved. The false alarm of the violation behavior is reduced, and the actual use effect is improved.

Description

Technical Field

[0001] The present invention relates to the field of AI image recognition technology, and in particular to the development and application of a spatiotemporal intelligent collaborative on-site operation safety risk prevention and control capability enhancement and application. Background Art

[0002] With the acceleration of industrialization and urbanization, safety risks in on-site operations are becoming increasingly prominent. This is particularly true in high-risk industries like construction, mining, and petrochemicals, where accidents cause economic and human losses to businesses and society. Conventional safety risk control methods typically rely on extensive manual intervention or use relatively common object detection algorithms to monitor on-site construction activities.

[0003] In the current field of safety supervision, general target detection algorithms are generally used to directly extract frames from video streams. Although this can reduce manpower investment, general target detection algorithms are only suitable for monitoring simple static images. The depth of violation identification in simple static images depends on the accuracy of the target detection model. In complex scenarios where workers and equipment interact frequently and multiple behaviors contain a sequence and a certain causal relationship, due to the low accuracy of the recognition model, false alarms often occur, which affects on-site construction.

[0004] Therefore, how to design a spatiotemporal intelligent collaborative on-site operation safety risk prevention and control capability improvement and application research and development that can solve the above technical problems. Summary of the Invention

[0005] In order to solve the above problems, the purpose of the present invention is to provide a spatiotemporal intelligent collaborative on-site operation safety risk prevention and control capability improvement and application research and development. The present invention combines time and space information to identify and track target objects in the video stream for real-time analysis. Due to the combination of multi-frame recognition results, it can greatly alleviate the situation of false detection of target objects, reduce false alarms of violations, and improve the actual use effect.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: Figure 1 As shown, in this embodiment, a method for improving the safety risk prevention and control capabilities of field operations and application development based on spatiotemporal intelligent collaboration is provided, including the following steps: Step S1: Configure the camera information actually used according to the site requirements; Step S2: Provide an object detection model designed according to regulatory requirements based on business needs; Step S3: According to the detection requirements, set an independent focus target name for each camera to perform real-time tracking combining algorithm service and ontology service; Step S4: Combine the target tracking algorithm of Deepsort and the independently started RTSP service to draw the moving routes of multiple moving target objects in the camera image, and combine the continuous multi-frame information to make early warning judgments on illegal behaviors; Step S5: According to the above-mentioned violation warning judgment result, according to the corresponding timestamp information, the corresponding original video stream image is searched and the warning image is drawn.

[0007] Furthermore, the specific process of step S3 is as follows: During the real-time tracking phase, the algorithm service is responsible for connecting to multiple cameras, providing corresponding recognition services, and pushing the recognition results and corresponding timestamp information to the Redis message queue for interaction with the ontology service. The ontology service controls the rotation of the camera and promptly receives and processes the recognition results provided by the algorithm service to verify related violations in combination with business logic.

[0008] Furthermore, the ontology service is specifically as follows: Step A1: Automatically take over camera permissions and set a regular inspection time based on the camera configuration information. The camera's regular inspection method is: each time it rotates a fixed angle, pauses for a fixed time to allow the camera to focus, and repeats the action until a 360-degree horizontal rotation is completed. Step A2: Count the object information detected by the camera during the first inspection. If the target object detected during the first inspection triggers the alarm condition, the camera is directly triggered to conduct a second inspection. If the pre-set target object is not detected during the first training, the camera is directly positioned at the angle to the ground operator label for routine operation monitoring.

[0009] Furthermore, in step A2, the re-detection steps are as follows: According to the statistical message results of the first inspection, the camera is directly rotated to the angle position of the target object to detect the special target object found during the inspection in order to monitor the target object; if the target object does not exist at this time, the camera is controlled to rotate, and at the same time, according to the detection results pushed by the algorithm service, Redis messages are consumed in real time until the target object is detected. The camera stops rotating to continuously focus on the target object for special operation supervision.

[0010] Furthermore, in step A2, the routine operation monitoring steps are as follows: First, focus the target object on the ground operator label, count the number of expressions of the ground operator, and obtain the camera image when the corresponding number is the highest; Secondly, based on the coordinates of the ground worker in the above picture and the resolution of the camera, the Euclidean distance between the centroid of each target object and the centroid of the picture is calculated; Finally, the mean of the Euclidean distance results is calculated to adjust the camera rotation distance to ensure that the operator is included in the camera image.

[0011] Furthermore, the specific steps for warning and judging the violation behavior are as follows: The ontology service consumes Redis messages and uses business logic rules to determine violations.

[0012] Furthermore, the algorithm service also includes pushing out the real-time detection video stream and saving the original video stream image in real time, detecting the face change, so that the subsequent main body service can call it.

[0013] The present invention has the following beneficial effects: 1- This invention uses a deep learning-based target detection algorithm, combined with corresponding multi-target tracking algorithms and other technologies, to achieve automatic control of the camera, focusing on the management and control of violations and dangerous behaviors in on-site operations.

[0014] 2- The present invention accesses the control of the camera and combines the results of the AI ​​intelligent algorithm model to conduct adaptive dynamic supervision of high-risk targets and violations, and realizes intelligent violation video synthesis of the corresponding results, facilitating real-time alarms in construction scenarios and subsequent problem tracing operations.

[0015] 3-The present invention mainly combines target detection and tracking algorithms, and accesses the inspection operation of the camera to automatically locate and inspect the construction site operation scenes, and conduct dynamic and intelligent supervision of typical illegal operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This invention illustrates Figure 1 ; Figure 2 This invention illustrates Figure 2 . DETAILED DESCRIPTION

[0017] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments: refer to Figure 1-2 As shown, in this embodiment, a method for improving the safety risk prevention and control capabilities of field operations and application development based on spatiotemporal intelligent collaboration is provided, including the following steps: Step S1: Configure the camera information to be used, such as the URL address, according to the site requirements; use Python code to access the network camera; Step S2: Provide an object detection model designed according to regulatory requirements based on business needs; Step S3: According to the detection requirements, set an independent focus target name for each camera to perform real-time tracking combining algorithm service and ontology service; Step S4: Combined with the target tracking algorithm of deepsort and the independently started RTSP service, the moving routes of multiple moving target objects in the camera image are drawn, and at the same time, combined with continuous multi-frame information, early warning judgment of illegal behaviors; such as throwing objects from high places, working without standardized clothing, people falling, etc., is carried out to reduce false detections caused by single-frame judgments, and at the same time, time series information is added for comprehensive judgment to improve the recognition accuracy of the algorithm service.

[0018] Step S5: Based on the violation warning results, the corresponding original video stream is searched for and the warning image is drawn based on the corresponding timestamp information. Optimally, all detected images within a certain time limit (e.g., 1 minute) are backtracked and synthesized into a video file. To reduce disk space usage, the algorithm service will automatically delete images periodically.

[0019] Furthermore, the specific process of step S3 is as follows: During the real-time tracking phase, the algorithm service is responsible for connecting to multiple cameras, providing corresponding recognition services, and pushing the recognition results and corresponding timestamp information to the Redis message queue for interaction with the ontology service. The ontology service controls the rotation of the camera and promptly receives and processes the recognition results provided by the algorithm service to verify related violations in combination with business logic.

[0020] Furthermore, the ontology service is specifically as follows: Step A1: According to the set camera configuration information, automatically take over the camera permissions and set the regular inspection time; the camera regular inspection method is: each time it rotates a fixed angle, pauses for a fixed time to allow the camera to focus, and repeats the action until it completes a horizontal 360-degree rotation; specifically, at the beginning of the first inspection, first rotate the camera clockwise, and stop for 2-3 seconds every 30 degrees to allow the camera to focus and obtain a clear picture. Detect the target object information involved in the current picture and push it to the redis message queue. Before each rotation is executed, the main body service will consume all the messages of the current angle picture, associate them with the angle information of the current camera, and save them in other message data. When the camera rotates one circle, the data detected in this rotation picture is counted. If no key target objects are found, such as the person in charge of the work, the fence of the work area, etc., the corresponding alarm information will be pushed.

[0021] Step A2: Count the object information detected by the camera during the first inspection. If the target object detected during the first inspection triggers the alarm condition, the camera is directly triggered to conduct a second inspection. If the pre-set target object is not detected during the first training, the camera is directly positioned at the angle to the ground operator label for routine operation monitoring.

[0022] Furthermore, in step A2, the re-detection steps are as follows: Based on the statistical results of the first inspection, the camera is directly rotated to the target object's angle to detect and monitor the special target object found during the inspection. If the target object is not present, the camera is controlled to rotate while consuming Redis messages in real time based on the detection results pushed by the algorithm service until the target object is detected. (Once the target object is detected, the camera stops rotating to maintain focus on the target object for special operation monitoring. During the first inspection, a full inspection cycle is still required regardless of whether the target object is detected.) Furthermore, after detecting the target object, the object detection frame's size is calculated as a percentage of the image. If the object size percentage is less than a certain threshold (such as 0.1), the focus is increased to magnify the target, ensuring that the object's size in the image is appropriately proportioned. This helps improve the model's detection accuracy and reduce false positives.

[0023] Furthermore, in step A2, the routine operation monitoring steps are as follows: First, focus the target object on the ground operator label, count the number of expressions of the ground operator, and obtain the camera image when the corresponding number is the highest; Secondly, based on the coordinates of the ground worker in the above picture and the resolution of the camera, the Euclidean distance between the centroid of each target object and the centroid of the picture is calculated; Finally, the mean of the above Euclidean distance results is calculated to adjust the camera rotation distance to ensure that as many ground workers as possible are included in the picture.

[0024] Furthermore, the specific steps for warning and judging the violation behavior are as follows: The ontology service consumes Redis messages and uses business logic rules to determine violations.

[0025] Furthermore, the algorithm service also includes pushing out the real-time detection video stream and saving the original video stream image in real time, detecting the face change, so that the subsequent main body service can call it.

[0026] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0027] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0028] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0029] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0030] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other manner. Any person skilled in the art may utilize the above-disclosed technical content to modify or modify the present invention into equivalent embodiments. However, any simple modifications, equivalent variations, and modifications to the above embodiments that do not depart from the technical content of the present invention and are based on the technical essence of the present invention remain within the scope of protection of the present invention.

Claims

1. A spatiotemporal intelligent collaborative on-site operation safety risk prevention and control capability improvement and application research and development, characterized by: The following steps are involved: Step S1: Configure the camera information actually used according to the site requirements; Step S2: Provide an object detection model designed according to regulatory requirements based on business needs; Step S3: According to the detection requirements, set an independent focus target name for each camera to perform real-time tracking combining algorithm service and ontology service; Step S4: Combine the target tracking algorithm of Deepsort and the independently started RTSP service to draw the moving routes of multiple moving target objects in the camera image, and combine the continuous multi-frame information to make early warning judgments on illegal behaviors; Step S5: According to the above-mentioned violation warning judgment result, according to the corresponding timestamp information, the corresponding original video stream image is searched and the warning image is drawn.

2. The method of improving the safety risk prevention and control capabilities of field operations and its application development based on spatiotemporal intelligent collaboration according to claim 1 is characterized in that: The specific process of step S3 is as follows: In the real-time tracking phase, the algorithm service is responsible for connecting to multiple cameras, providing corresponding recognition services, and pushing the recognition results and corresponding timestamp information to the Redis message queue for interaction with the ontology service; The camera's rotation is controlled through the ontology service and the recognition results provided by the processing algorithm service are received in a timely manner to verify the relevant violations in combination with the business logic.

3. The method of improving the safety risk prevention and control capabilities of field operations and its application development based on spatiotemporal intelligent collaboration according to claim 2 is characterized in that: The ontology services are as follows: Step A1: Automatically take over camera permissions and set a regular inspection time based on the camera configuration information. The camera's regular inspection method is: each time it rotates a fixed angle, pauses for a fixed time to allow the camera to focus, and repeats the action until a 360-degree horizontal rotation is completed. Step A2: Count the object information detected by the camera during the first inspection. If the target object detected during the first inspection triggers the alarm condition, the camera is directly triggered to conduct a second inspection. If the pre-set target object is not detected during the first training, the camera is directly positioned at the angle to the ground operator label for routine operation monitoring.

4. The method of improving the safety risk prevention and control capabilities of field operations and its application development based on spatiotemporal intelligent collaboration according to claim 3 is characterized in that: In step A2, the second inspection steps are as follows: According to the statistical information results of the first inspection, the camera is directly rotated to the angle position of the target object to detect the special target object found during the inspection to monitor the target object; If the target object does not exist at this time, the camera is controlled to rotate. At the same time, according to the detection results pushed by the algorithm service, Redis messages are consumed in real time until the target object is detected. The camera stops rotating to continuously focus on the target object for special operation supervision.

5. The method of improving the safety risk prevention and control capabilities of field operations and its application development based on spatiotemporal intelligent collaboration according to claim 3 is characterized in that: In step A2, the routine operation monitoring steps are as follows: First, focus the target object on the ground operator label, count the number of expressions of the ground operator, and obtain the camera image when the corresponding number is the highest; Secondly, based on the coordinates of the ground worker in the above picture and the resolution of the camera, the Euclidean distance between the centroid of each target object and the centroid of the picture is calculated; Finally, the mean of the Euclidean distance is calculated and the camera rotation distance is adjusted to ensure that the operator is included in the camera image.

6. The method of claim 1 for improving the safety risk prevention and control capabilities of field operations using spatiotemporal intelligent collaboration and its application development is characterized by: The specific steps for warning and judging the violation behavior are as follows: The ontology service consumes Redis messages and uses business logic rules to determine violations.

7. The method of claim 1 for improving the safety risk prevention and control capabilities of field operations using spatiotemporal intelligent collaboration and its application development is characterized by: The algorithm service also includes pushing out the real-time detection video stream and saving the original video stream image and the detection image in real time so that the subsequent ontology service can call it.

Images