Study On Algorithm For Camouflage Target Detection And Covering Target Inpainting In Mine

Coal as the primary resource is demanded largely all the time. Because of seriousenvironment and technique limitation, at present, rail transport is the main method tooutput coal in most mine. Monocular vision alarming system is developed for the goalof rail recognition, miner detection and ranging. This system make up somedisadvantages of traditional rail transport to some degree reducing accidents like rearcollision and cracking miners and improving productivity security. Owing to thecomplicated mine environment and image processing pertinence, this system can’textract rail and miners for the limitation of camouflage and lights shielding.Based on the shortcomings listed above, this paper presents the development of themoving target detection and image restoration, and compare kinds of algorithms throughtheory and advantages. Aimed at camouflage and light shielding of miners in thepractical mine image, two methods are proposed, one is based on color and texture, theother on texture restoration priority. The primary research includes:①Aimed at the minors camouflage, traditional methods shows dilemma to detectmines. Through the analysis of present theories, a method based on color and texturemode is proposed. By means of combining Gaussian Mixture Model (GMM) for colormode and Local Binary Pattern (LBP) for texture mode, finally, the minors can bedetected accurately.②Aimed at big error on detecting minors when employing light area in replace oflocal light shielding, we first erase the shielding light then detect minors. Through thecomparison with traditional theories of image restoration, a improved algorithm basedon priority is proposed to erase light shielding. Experiment results show that theimproved method is effective on minors detection.