Online monitoring system based on intelligent violation behavior recognition device

Abstract

The invention discloses an on-line monitoring system based on an intelligent violation behavior recognition device, and particularly relates to the field of oil field on-line monitoring. A working condition video AI early warning module is used for carrying out video information acquisition on an above-well construction site and underground equipment and personnel in an oil field; the method comprises the following steps of: transmitting a video to a multi-scale target AI detection and analysis module to construct a multi-scale feature expression and a detection model, comparing and analyzing data analyzed by the video multi-scale target AI detection and analysis module by a dynamic environment video AI analysis module so as to effectively classify the data, and performing violation identification through a violation behavior intelligent identification module in the whole process. The analysis efficiency and the acquisition efficiency are improved through cooperation of the video AI cloud collaborative analysis module, and the algorithm precision is improved through the video AI model automatic training module, so that the violation recognition precision is improved, violation behaviors in the overall environment in an oil field can be intelligently recognized, and the safety of personnel and the work normalization are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil field online monitoring, and more specifically, to an online monitoring system based on an intelligent identification device for illegal behavior. Background Art

[0002] With the rapid development of the national economy and the continuous acceleration of urbanization, my country's demand for energy is increasing. As the pillar of energy, the oil and gas industry is related to the country's economic lifeline. It is of great significance to carry out relevant technical research on oil and gas exploration and development.

[0003] In recent years, with the continuous expansion of the scope of oil and natural gas development, there will be certain unsafe behaviors of personnel, unsafe conditions of objects and adverse environmental factors in the process of oil and natural gas exploitation. By adopting technologies such as multi-scale target detection, intelligent recognition of illegal behaviors, lightweight compression of models and autonomous training of analysis models, 21 types of target detection such as research and development personnel, vehicles, animals, etc., and 11 illegal behaviors such as incomplete labor protection wear, no safety rope for high-altitude operations, and people standing under hanging objects, in order to reduce violations and improve work standardization and safety, an online monitoring system based on an intelligent recognition device for illegal behaviors is proposed. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides an online monitoring system based on an intelligent identification device for traffic violation behavior to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: an online monitoring system based on an intelligent identification device for traffic violation behaviors, comprising an operating condition video AI early warning module, a multi-scale target AI detection and analysis module, a dynamic environment video AI analysis module, an intelligent identification module for traffic violation behaviors, a video AI cloud collaborative analysis module and a video AI model automatic training module, wherein the operating condition video AI early warning module is connected and coordinated with the multi-scale target AI detection and analysis module, the multi-scale target AI detection and analysis module is connected and coordinated with the dynamic environment video AI analysis module, the dynamic environment video AI analysis module is connected and coordinated with the intelligent identification module for traffic violation behaviors, the intelligent identification module for traffic violation behaviors is connected and coordinated with the video AI cloud collaborative analysis module, and the video AI cloud collaborative analysis module is connected and coordinated with the video AI model automatic training module.

[0006] Preferably, the working condition video AI warning module includes a static target recognition unit, a dynamic target recognition unit, a behavioral target recognition unit and a logical event recognition unit.

[0007] Preferably, the multi-scale target AI detection and analysis module is used to construct a multi-scale feature expression, adopt an FPN network structure, obtain feature maps of different scales, and then use a top-down and lateral connection method to construct a detection model to achieve real-time detection of target objects in the picture.

[0008] Preferably, the dynamic environment video AI analysis module is used to cut the target object image into a large number of pixel points for labeling, simulate the neural network to establish a data calculation model, send the labeled target object into the data model for training, and generate a target object model.

[0009] Preferably, the dynamic environment video AI analysis module includes a moving object recognition unit and a fireworks and oil leak recognition unit.

[0010] Preferably, the traffic violation intelligent recognition module includes a fusion human body recognition unit and a tracking unit.

[0011] Preferably, the frequency AI cloud collaborative analysis module includes a model stacking unit and a neural architecture search unit.

[0012] Preferably, the video AI model automatic training module includes an application data labeling unit and a deep learning unit.

[0013] Technical effects and advantages of the present invention:

[0014] The present invention discloses an online monitoring system based on an intelligent identification device for illegal behaviors. The system collects video information of an on-well construction site, underground equipment and personnel in an oil field through a working condition video AI early warning module, and transmits the video information to a multi-scale target AI detection and analysis module to construct a multi-scale feature expression. The FPN network structure is adopted to obtain different feature mapping data, thereby constructing a detection model, and performing real-time detection and analysis on the data collected by the video AI early warning module. The dynamic environment video AI analysis module compares and analyzes the data analyzed by the video multi-scale target AI detection and analysis module, thereby performing effective classification. The classified data is transmitted to the intelligent identification module for illegal behaviors for illegal identification. In the whole process, the cooperation with the video AI cloud collaborative analysis module reduces the cloud computing amount and network bandwidth resource occupation, realizes the "cloud-edge-end" collaborative sharing of intelligent algorithm analysis resources, improves the analysis efficiency and collection efficiency, and improves the algorithm accuracy of the video AI model automatic training module, thereby improving the accuracy of illegal identification, so that illegal behaviors in the overall environment of the oil field can be intelligently identified, so as to improve the safety of personnel and the standardization of work. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the flow of the online monitoring system of the intelligent identification device for traffic violation of the present invention.

[0016] Figure 2This is a schematic diagram of the video AI multi-scale target detection analysis principle of the present invention.

[0017] Figure 3 This is a technical architecture diagram of the AI ​​intelligent analysis of videos in a dynamic environment of the present invention.

[0018] Figure 4 This is a diagram showing the effect of video AI detection and analysis in a dynamic environment according to the present invention.

[0019] Figure 5 This is a schematic diagram of the technical principle of intelligent identification of traffic violation behavior of the present invention.

[0020] Figure 6 This is a schematic diagram of the video AI cloud collaborative analysis technology of the present invention.

[0021] Figure 7 This is a schematic diagram of the video AI model autonomous training technology of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0023] Example 1

[0024] like Figure 1 As shown, an online monitoring system based on a traffic violation intelligent recognition device includes an operating condition video AI early warning module, a multi-scale target AI detection and analysis module, a dynamic environment video AI analysis module, a traffic violation intelligent recognition module, a video AI cloud collaborative analysis module and a video AI model automatic training module, the operating condition video AI early warning module is connected and coordinated with the multi-scale target AI detection and analysis module, the multi-scale target AI detection and analysis module is connected and coordinated with the dynamic environment video AI analysis module, the dynamic environment video AI analysis module is connected and coordinated with the traffic violation intelligent recognition module, the traffic violation intelligent recognition module is connected and coordinated with the video AI cloud collaborative analysis module, and the video AI cloud collaborative analysis module is connected and coordinated with the video AI model automatic training module.

[0025] Through the working condition video AI early warning module, video information of the well construction site, underground equipment and personnel in the oil field is collected and transmitted to the multi-scale target AI detection and analysis module to construct a multi-scale feature expression. The FPN network structure is used to obtain different feature mapping data, thereby constructing a detection model. The data collected by the video AI early warning module is detected and analyzed in real time, while the dynamic environment video AI analysis module compares and analyzes the data analyzed by the video multi-scale target AI detection and analysis module for effective classification. The classified data is transmitted to the violation behavior intelligent recognition module for violation identification. In the whole process, the video AI cloud collaborative analysis module is used to reduce the cloud computing workload and network bandwidth resource usage, realize the "cloud-edge" collaborative sharing of intelligent algorithm analysis resources, improve analysis efficiency and collection efficiency, and the video AI model automatic training module improves algorithm accuracy, thereby improving the accuracy of violation identification.

[0026] Example 2

[0027] The working condition video AI warning module includes a static target recognition unit, a dynamic target recognition unit, a behavioral target recognition unit and a logical event recognition unit.

[0028] The static target recognition unit can be used to identify the uniforms worn by personnel, the model and license plate of vehicles, identify and monitor various areas such as warning areas, danger areas and restricted areas, and identify the pumping units, pipelines, booms, escape signs, etc. in the equipment;

[0029] The dynamic target recognition unit can be used to identify the protective gear worn by personnel, identify and record the location, number and retention time of vehicles, identify and record the number, location and retention time of animals, identify wellhead leakage, ground leakage, pipeline leakage, valve dripping, etc., identify the pumping unit shutdown, oil and water tank overflow, etc. in the equipment, and identify the occurrence of smoke and fire;

[0030] Through the behavioral target recognition unit, it can identify people smoking, making phone calls, and gatherings; identify vehicles wandering, the presence or absence of drivers, and illegal parking; identify the forward and reverse rotation of oil pumps, the fall of hot work equipment, etc.; and identify operations such as throwing balls, cleaning, and adding drugs during operations;

[0031] Through logical event recognition, it is possible to identify whether personnel are on patrol, off duty, sleeping on duty, etc., whether the vehicle has a driver and whether it has parked illegally, the rationality of the escape signs in the equipment and the road crossing protection situation are identified, water leakage in the valve room and the value of the valve room instrument table is reset to zero, and it is possible to identify whether the spacing between hot work equipment is insufficient during the operation, whether the fire-fighting equipment is not detected, whether the safety rope is not tied during high-altitude operation, whether there are people standing under the hanging objects, whether the crane support legs are not extended, the excavation soil pile is too high, and the pit is too shallow.

[0032] Example 3

[0033] like Figure 2 As shown, the multi-scale target AI detection and analysis module is used to construct a multi-scale feature expression. After obtaining feature maps of different scales using the FPN network structure, a detection model is constructed using a top-down and lateral connection method to achieve real-time detection of target objects in the picture, so as to avoid the situation in which there are many construction workers and various illegal behaviors during underground operations and ground engineering construction, and a single target detection technology is difficult to accurately identify illegal behaviors.

[0034] Example 4

[0035] like Figure 3 and Figure 4 As shown, the dynamic environment video AI analysis module is used to cut the target image into a large number of pixel points for labeling, simulate the neural network to establish a data calculation model, send the labeled target object into the data model for training, and generate a target object model.

[0036] The target object image is cut into a large number of pixel points for annotation, and a neural network is simulated to establish a data calculation model. The labeled target object is sent to the data model for training to generate a target object model. By comparing and analyzing the collected images with the neural network model, accurate identification, effective classification, real-time processing, traceability analysis, and automatic alarm push are achieved for seven major types of abnormal conditions, including personnel, vehicle volume, fireworks, animals, oil leakage, equipment, and action status at the production site.

[0037] The dynamic environment video AI analysis module includes a moving object recognition unit and a fireworks and oil leak recognition unit.

[0038] Mobile object recognition unit: Based on deep learning edge detection and recognition technology, through data collection, data preprocessing, target information labeling, and building a neural network model, it can ultimately achieve the recognition of various targets with fixed edge contours such as people, cars, animals, and equipment.

[0039] Fireworks and oil leak identification unit: The pixel values ​​of the two pictures collected successively are interpolated. If the interpolation value is not zero, the changed parts are cut out, and the presence of fireworks is judged based on the color of the fire and other characteristics. The collected images are compared and analyzed with the established neural network models of morning glow, evening glow and car lights to eliminate false alarms.

[0040] Example 5

[0041] like Figure 5 As shown, the traffic violation intelligent recognition module includes a fusion human body recognition unit and a tracking unit.

[0042] By integrating human behavior recognition and tracking methods, the position changes of the target object's skeleton points or key points are obtained to capture the target object's posture changes, and a violation behavior recognition model is constructed to achieve intelligent recognition of violations at the construction site. Target tracking technology is used to predict the movement trajectory of the target object, and the relationship between video time series is used to establish the connection between video frames to solve the occlusion problem.

[0043] Example 6

[0044] like Figure 6 As shown, the frequency AI cloud collaborative analysis module includes a model stacking unit and a neural architecture search unit.

[0045] Through model stacking and neural architecture search technology, multiple models are encapsulated into a standardized algorithm model warehouse and sent to the edge to complete computational analysis, reducing the amount of cloud computing and network bandwidth resources occupied, realizing the "cloud-edge" collaborative sharing of intelligent algorithm analysis resources, and improving analysis efficiency and collection efficiency.

[0046] Example 7

[0047] like Figure 7 As shown, the video AI model automatic training module includes an application data labeling unit and a deep learning unit.

[0048] By applying methods such as data annotation and deep learning, we build model prediction, reasoning, and training environments and the oil field’s own sample library to achieve autonomous training and optimization iteration of AI algorithms, and continuously improve algorithm accuracy.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An online monitoring system based on an intelligent identification device for illegal behavior, Features: It includes a working condition video AI early warning module, a multi-scale target AI detection and analysis module, a dynamic environment video AI analysis module, a violation behavior intelligent identification module, a video AI cloud collaborative analysis module and a video AI model automatic training module. The working condition video AI early warning module is connected and coordinated with the multi-scale target AI detection and analysis module, the multi-scale target AI detection and analysis module is connected and coordinated with the dynamic environment video AI analysis module, the dynamic environment video AI analysis module is connected and coordinated with the violation behavior intelligent identification module, the violation behavior intelligent identification module is connected and coordinated with the video AI cloud collaborative analysis module, and the video AI cloud collaborative analysis module is connected and coordinated with the video AI model automatic training module.

2. According to claim 1, an online monitoring system based on an intelligent identification device for illegal behavior, Features: The working condition video AI warning module includes a static target recognition unit, a dynamic target recognition unit, a behavioral target recognition unit and a logical event recognition unit.

3. According to claim 1, an online monitoring system based on an intelligent identification device for illegal behavior, Features: The multi-scale target AI detection and analysis module is used to construct a multi-scale feature expression. After obtaining feature maps of different scales using the FPN network structure, a detection model is constructed using a top-down and lateral connection method to achieve real-time detection of target objects in the picture.

4. According to claim 1, an online monitoring system based on an intelligent identification device for illegal behavior, Features: The dynamic environment video AI analysis module is used to cut the target object image into a large number of pixel points for labeling, simulate the neural network to establish a data calculation model, send the labeled target object into the data model for training, and generate a target object model.

5. According to claim 1, an online monitoring system based on an intelligent identification device for illegal behavior, Features: The dynamic environment video AI analysis module includes a moving object recognition unit and a fireworks and oil leak recognition unit.

6. An online monitoring system based on an intelligent identification device for traffic violation according to claim 1, Features: The traffic violation intelligent recognition module includes a fusion human body recognition unit and a tracking unit.

7. An online monitoring system based on an intelligent identification device for traffic violation according to claim 1, Features: The frequency AI cloud collaborative analysis module includes a model stacking unit and a neural architecture search unit.

8. An online monitoring system based on an intelligent identification device for traffic violation according to claim 1, Features: The video AI model automatic training module includes an application data labeling unit and a deep learning unit.

Images