Improved Retinex-Net Algorithm for Low Quality Image Enhancement in Open-pit Mine
Low-quality image enhancement is an important part of the perception system of unmanned wide-body vehicle in open-pit mine.The image acquired by monocular camera is susceptible to many factors such as dust,rain,snow and fog in mining area and violent vibration.Aiming at the problems of high noise and color distortion in the traditional image enhance-ment algorithm,an improved Retinex-Net algorithm is proposed to enhance the image of open-pit mine.Cyclic adversarial gen-eration network and two-channel residual network are used to improve the enhancement and denoising parts.The cyclic adver-sarial generation network generates more natural and realistic enhanced results by learning the mapping relationship between low-light and normal-light images.The dual-channel residual network can effectively remove noise and artifacts in low-light im-ages by processing brightness and chrominance information simultaneously.The experimental results show that the proposed method is superior to the existing methods in both objective and subjective evaluation indexes.The proposed Retinex-Net algo-rithm provides an effective scheme to solve the image quality problem in open-pit mine.
NETL
NSTL
万方数据
