Research On Coal Heap Smoke Detection Algorithm Based On Convolution Neural Network

With social and economic development,coal plays an important role in the field of economic production.The long-term accumulation of coal is prone to spontaneous combustion,which will not only cause losses to the economic value of coal,but also cause major safety accidents.To solve this problem,a smoke detection algorithm based on surveillance video is proposed.The smoke produced by spontaneous combustion of coal pile is detected and alerted by collecting field information from the camera of port yard.two smoke detection algorithms are designed in this paper.This paper studies the algorithm from the following aspects.(1)A smoke detection algorithm based on spatiotemporal characteristics and deep learning is designed,which is improved in two aspects.On the one hand,the algorithm based on spatiotemporal feature is used to extract and screen the image of movin g target region,then the single frame image is screened out and the key surface image is extracted by using the spatiotemporal feature of video.The experimental results show that the improved algorithm improves the speed greatly.On the other hand,the network layer based on AlexNet deep classification network is improved,and the original Softmax classification layer is replaced by SVM classification layer in the network structure.(2)The deep learning algorithm module replaces the steps of feature extr action and classification learning,which can achieve high accuracy.But this method is still the traditional smoke detection framework.therefore,a YOLO v3 smoke detection framework based on Vi Be extraction of moving regions is designed.The Vi Be algorit hm can effectively frame the area of interest of the dynamic target and extract the smoke foreground.On this basis,the smoke data set can be trained by YOLO v3 to detect the smoke region.Experimental results show that the proposed algorithm can effectiv ely detect the smoke in coal piles.Finally,aiming at the application of video smoke detection algorithm in practice,a method of smoke video statistical testing is designed to calculate the accuracy rate and recall rate of the test video.In order to objectively evaluate the algorithm in this paper,it is compared with the mainstream smoke detection algorithm.the experimental results verify the effectiveness of this algorithm.