Low-light Image Enhancement Based On Retinex And Camera Response Mechanism In Coal Mines

Under the background of smart mine construction,the intelligent coal mining technology featuring unmanned and intelligent is gradually coming into practice.The complex environmental factors in underground coal mines lead to problems such as low illumination and high noise in the collected images,which hinder the popularization and application of intelligent detection and intelligent perception technologies with image processing as the core.In response to the above problems,research on enhancement techniques such as brightness enhancement,noise suppression,and color restoration of underground low-light images was carried out.A low-light image enhancement algorithm for coal mine underground with improved depth Retinex and camera response fusion is proposed.Firstly,the improved depth Retinex-Net algorithm with decomposition,denoising and restoration-enhancement effects is proposed for the problems of high noise,color distortion and blurred details of Retinex-Net.After the effectiveness of the method is tested by ablation experiments and comparison experiments on classical datasets,the algorithm is used to process the original image and obtain the initial enhanced image;then,the Then,the camera response mechanism is used to simulate the exposure process and find the optimal exposure rate to generate the virtual enhanced image;finally,the initial enhanced image and the virtual enhanced image are reconstructed and fused using multi-scale image fusion to obtain the final enhanced image.Comparative experiments are conducted on the LOL dataset,and the results show that the method in the paper improves 20.5%,5.1%,2.4% and 12.5% over the Retinex-Net algorithm in PSNR,SSIM,UQI and FSIM metrics,respectively.The captured underground low-illumination images with different scenes from Wangpo coal mine in Tiandi,Shanxi and Baode coal mine in Shenhua,China were enhanced by the proposed algorithm,and the enhancement effect was analyzed by comparing with six algorithms,including MSRCR,LIME,SRIE,Retinex-Net,R2 RNet and RRDNet,in terms of information entropy,contrast-enhanced image evaluation and The proposed algorithm achieves good performance in three indexes of natural image quality evaluation,and the qualitative and histogram analysis results show that the algorithm can effectively suppress halos,artifacts and noise in images and improve image quality to meet the requirements of image quality for intelligent detection and intelligent sensing technology based on downhole image processing.The paper has 30 figures,10 tables and 74 references.