Design Of Miner's Security Wearable Device Detection System Based On Deep Learning

Mineral resources provide an important material foundation for people's production and life,and the rapid development of the national economy is also inseparable from sufficient energy.However,according to the survey results,accidents have occurred frequently in mining in recent years.In order to protect the life and safety of front-line miners and to prevent the miners from being injured during the mining process,mines are generally wearing security device before they can mine.However,due to the complex environment,high mobility of personnel,and low awareness of safe mining among frontline miners,it is often difficult for supervisors to supervise in real time whether each miner is wearing full security equipment.In order to solve the problem of real-time supervision,this paper proposes a method that uses deep learning technology to detect whether the security equipment of miners is wearing,and develops a set of SSD-MobileNet-based security monitoring equipment for miners.Through such an automated and intelligent detection system,the complexity and repetitive problems of miner security wearable device detection can be effectively solved.Further effectively restrict the safe production behavior of miners and realize the intelligentization of the mine safety management system.The main work of this article is as follows:(1)This paper proposes the overall design scheme of the security wearable device detection system for miners' personnel.The detection system consists of hardware such as pressure switch,PLC,camera,host computer and gate system.The system has functions such as dynamic sensing,data acquisition,target detection,data statistics,gate control and human-computer interaction.(2)This article has carried out research on the theoretical knowledge of deep learning.The development process,network structure and training process of convolutional neural network are described,and the calculation process of forward propagation is introduced with a typical CNN network structure.Then,this paper studies the target detection algorithms of security wearables of miners based on deep learning,compares the working principle and network structure of two-step detection algorithm and single-step detection algorithm,and analyzes the advantages and disadvantages of the two algorithms.(3)This paper has carried out experimental verification on three major target detection models.A data set of miners' security wearable devices was first prepared,and then three different detection models of Faster R-CNN,YOLO,and SSD-MobileNet were trained using this data set.Finally,the detection accuracy and speed of the three models were compared.Judging from the results,the detection accuracy and speed of the SSD-MobileNet algorithm model both meet the requirements of the detection system design.Therefore,this article will use this model to build a detection system.After actual verification,the system has good stability and anti-interference,and can meet the needs of rapid detection of security wearable devices for miners.(4)This article completes the specific implementation of the detection system.This includes the hardware selection of the inspection system,the design and installation of each sub-module of the inspection system,the layout of the upper computer interface of the inspection system and the preparation of the logic business,and the system construction,testing and optimization.