Font Size: a A A

Research On Joint Control System Of Mine Inclined Lane Based On AI Algorithm In Video Terminal

Posted on:2023-07-02Degree:MasterType:Thesis
Country:ChinaCandidate:X W QinFull Text:PDF
GTID:2531306815466014Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
At present,various coal mining enterprises have deployed industrial TV monitoring systems underground,but these video monitoring systems are relatively simple in design,with the cameras only undertaking the task of video capture,while the analysis and processing of image information is completed by the cloud server in the monitoring centre above the shaft.This has resulted in low real-time monitoring systems,high network resource usage and slow response to monitoring equipment linkage.To address the above problems,applying technologies such as image processing,deep learning and edge computing,it proposes a joint control system based on the AI(Artificial Intelligence)algorithm implemented at the video end of the mine slope,which implements the target detection task of images at a high-performance embedded terminal and improves the timeliness of the smart video surveillance system.Based on the research related to image enhancement,image target detection,video linkage and edge computing technologies,an architecture for a linked linkage control system for mine inclined shafts is designed,and the functions of each level of the system are divided based on edge cloud collaboration technology.Aiming at the problem of image enhancement in harsh downhole environments,a multi-scale Retinex algorithm based on fast weighted bootstrap filtering is introduced to sharpen the detailed features of images.To enable reliable deployment and effective running of the neural network model over embedded terminals,the lightweight version of YOLOv3,Tiny YOLOv3,is improved in the present study by constructing a downsampling residual module with depth-separable convolutional units as the main body,and designing an improved Tiny YOLOv3 network based on the downsampling residual module.This lightweight neural network enables fast and accurate identification of underground pedestrians on embedded terminals.In order to ensure that the mine’s underground inclined lane is " travelling without people and pedestrians without travelling",the video processing node and the communication node in the winch linkage control are developed in hardware,and the sound and light alarm device in the inclined lane as well as the cloud service platform in the monitoring centre are introduced.The validity of partial functions of the mine inclined lane linkage control system is verified through related functional tests.Testing and verification results demonstrate that proposed system can achieve exact identification of video targets at the high performance embedded terminal;compared to the original way of video data transmission to the ground server for processing,the direct processing at the video end of the present work,its system’s timeliness and reliability are guaranteed;during winch operation,if there is an illegal pedestrian in the monitoring area,the sound and light alarm device in the inclined lane implements linkage alarm.Research work in this thesis provides foundation for the realization of surveillance of mine video in real time and response of equipment in real time,which guarantees the security of coal mine production.Figure [50] Table [9] Reference [79]...
Keywords/Search Tags:video linkage control, edge cloud collaborative technology, multi-scale Retinex, image recognition, tiny YOLOv3
PDF Full Text Request
Related items