Identification And Research Of Safety Regulations And Conditions In Coal Mines Using Surveillance Videos

The working environment of coal mines is a complex and constantly changing model involving humans,machines,and the environment.It poses numerous hazardous factors such as many workers,complicated equipment,dispersed work areas,harsh environmental conditions,and difficult management.These factors not only threaten the safety of coal mine workers but also affect the production efficiency of mining enterprises.Traditional video surveillance systems mainly accomplish two tasks:evaluating responsibilities after accidents occur and relying on real-time monitoring by personnel.Therefore,traditional video surveillance systems have shortcomings such as the inability to issue real-time warnings and an inability to eliminate the effects of real-time monitoring on humans.In addition,existing object detection and human pose estimation technologies only focus on people or objects themselves and cannot obtain correlation information between people and objects,making they difficult to predict safety accidents caused by human factors.To address these two issues,this paper uses image processing,deep learning,and other algorithms to identify and interpret the safety standards and status of coal mine monitoring videos.Experimental results have shown that this method is effective.The research topic is based on the monitoring videos of coal mines,focusing on the methods for identifying and assessing compliance with safety regulations and the status of coal mines.The main contents are as follows:(1)Addressing factors such as low light,dust,and uneven lighting in coal mine work environments,this study analyz the processing effects of the Retinex algorithm,2D gamma function algorithm,and CLAHE algorithm for coal mine monitoring images and proposed a coal mine work video image enhancement algorithm with contrast limited adaptive retinex gamma equalization(CLARGE).First,the RGB image of the coal mine work video image is converted into an HSV image,and the brightness component is extracted.The Retinex algorithm separat the brightness component into reflection and illumination components.Then,the 2D gamma function is used to adjust the illumination value of each region,weakening regions with high illumination intensity and enhancing regions with low illumination intensity.Finally,adaptive contrast enhancement using CLAHE is applied to enhance image details while preserving original features.By comparing four different scenarios and using three evaluation criteria of information entropy,average gradient,and standard deviation,the 2D gamma function correction algorithm,CLAHE algorithm,and CLARGE algorithm are compared.The results show that the CLARGE algorithm effectively improv lighting unevenness and dust phenomenon and has significant advantages in improving image information,clarity,and contrast.(2)Due to the limited number of public datasets available for coal mining work environments,this study construct a MKZTAQ dataset reflecting worker safety status and clothing under coal conditions based on current public datasets’ characteristics and structures.Since some environmental pictures of coal mine operations have many contrasts,which may result in poor detection performance,this problem was addressed by converting safety wear detection into a segmentation identification problem based on safety wear.By analyzing features enhanced recognition and expanded perception,Res Net network extracts features,and DANet and Deep ASPP are used for feature segmentation training.The results showed that the two algorithms were effective in detecting safe wear in coal mining work environments,and DANet is found to have higher accuracy and intersection-over-union ratio than Deep ASPP through experimental verification.(3)This study select unmanned areas in the coal discharge area and operating personnel safety monitoring of belt conveyors in the coal mine work environment to identify limited area environmental safety status.To address violations caused by workers entering unmanned areas in the coal discharge area,the spatial topology model of the coal mine operation process is constructed based on the analysis of personnel’s security status within the designated area of the coal mining video.With YOLO v5 recognition algorithm,operator target boxes are extracted,and calculation is performed with the dangerous zone coordinate box to determine whether operators have entered the danger zone.In terms of judging the safety status of operating personnel on belt conveyor operations,stereoscopic environmental analysis is converted to stereoscopic projection analysis.A pixel topology model is constructed for the belt conveyor operation,Open Pose is used to extract the skeleton coordinates of personnel and build the skeleton coordinate pixel set,and the dangerous area pixel set of the operating area is matched to determine the dangerous status of operator behavior.The results show that both behavioral modeling models are effective in identifying worker danger and issuing warnings.This paper mainly explores the methods for identifying violation behaviors and statuses in coal mine working environment video surveillance.Through the verification of coal mining operation monitoring videos,it is demonstrated that the proposed method has a good recognition effect and can accurately identify violation behaviors and statuses.This research has certain significance for coal mine safety production and early warning of safety accidents in coal mine working environments.