Research And Application Of Computing Equilibrium Method For Mine Edge Visual Perception

Edge visual perception technology is currently a research hotspot in the field of coal mine safety monitoring.However,the complex environment of high temperature,high humidity and high dust in coal mines makes it difficult for the existing edge visual perception technology to be directly applied to coal mines.The main problems are as follows: First,the network and computing resources of edge computing equipment under coal mines are limited,which makes it difficult to balance the accuracy of the underground edge visual perception model with the hardware constraints of the equipment;Second,the high temperature,high humidity and high intrinsic safety requirements in coal mines lead to high temperature and unstable operation of edge computing equipment.In response to the above issues,this thesis combines neural architecture search and selective inference methods to study the balance method of mine edge visual perception calculation.The specific content includes:(1)Aiming at the problem that the accuracy of the visual perception model of the underground edge is difficult to balance with the hardware constraints of the equipment,a hypernet based on mixed depth convolution is constructed,a loss function under the historical architecture constraints is designed,and a generative architecture search method is proposed,which solves the problem of low accuracy caused by multi model forgetting and low efficiency of architecture search under different hardware constraints,and realizes the balance between the accuracy of the model and the hardware constraints of the equipment.(2)Aiming at the problem that downhole edge computing equipment is prone to unstable operation due to high temperature,a perceptual hash method based on accelerated up robust features(SURF)and singular value decomposition(SVD)is designed The problem of inefficient computing resource consumption of underground edge computing equipment is high,which realizes the stable and efficient operation of edge visual perception model on underground edge computing equipment.(3)On the basis of the above research,in response to the demand for low-cost and efficient construction of underground edge visual perception models,a low code underground edge visual perception development system has been developed,including modules such as data annotation,model construction,and model training.This solves the problem of high difficulty in developing underground edge visual perception technology and improves the practicality of edge visual perception technology in the underground.This thesis conducted experiments on public datasets and real datasets in coal mines to evaluate the proposed method,and the experimental results verified the effectiveness of the proposed mine edge visual perception calculation equalization method.