Research On Image Enhancement Of Algorithm And Detection Algorithm For Water Probe In Coal Mine Underground

Since convolutional neural network has been widely used in artificial intelligence devices,compared with traditional methods,the method based on deep learning has reached a quite high level of image processing ability.However,most of the current deep learning models use public paired data sets,which will lead to the weak generalization ability of the model in practical application,and the model trained by general data sets has poor enhancement effect in specific scenes.In the detection of underground objects in coal mines,there are problems of low detection accuracy and slow speed.Therefore,this paper takes improving the image enhancement results,detection accuracy and detection speed of the deep learning model as the breakthrough point,and studies the low-light image enhancement technology combined with generating countermeasure network,and detects the water probe with YOLOv7.This paper mainly optimizes and improves the Enlighten GAN network and YOLOv7 network to improve the enhancement effect and detection accuracy of the model.The main improvements of this paper are as follows:(1)At present,the low-illumination image enhancement methods are easy to lead to weak generalization ability and over-fitting phenomenon due to the lack of sufficient matching low-illumination data sets in real scenes.To solve this problem,this paper proposes an improved low-illumination image enhancement method based on hole convolution.On the basis of keeping the original Enlighten GAN network backbone unchanged,a series of improvements are made: firstly,the multi-branch hole convolution module is used to replace the conventional convolution module,and then the residual connection and channel attention module are introduced.(2)In order to solve the problem of low recognition accuracy of water probe caused by complex underground environment,uneven lighting and background interference in coal mine,this paper improves YOLOv7.Firstly,the improved Transition＿Block module is used to replace the deep network Transition＿Block module,and then the mixed attention mechanism is added before feature fusion to improve the sensitivity of the network to small-scale targets and reduce the impact of noise.On this basis,SIo U is used to replace CIo U in the original YOLOv7 network model to optimize the loss function,reduce the degree of freedom of the loss function and improve the network robustness.