Application Of Rough Set Theory In Coal Gangue Image Recognition

Coal is one of primary energy in our country, and the online recognition technology of coal and gangue is an advanced technology to realize the industrial automation of coal mine. Because the large amount of uncertain information is existed in coal and gangue image, and imprecise, disaccord and imperfect information can be analyzed and handled effectively by using the rough set theory, this thesis mainly uses rough set theory to preprocess and distinguish coal gangue image.The preprocessing of coal gangue image includes image filtering, image enhancement and edge detection. Firstly some basic concepts of rough set theory are introduced. According to the unresolved relation of rough set, the noise is divided and removed from the original coal gangue image. In order to enhance the edge, an improved median filtering method based on rough set to enhance coal gangue image is used in this paper, it makes edge and texture information of the image more obvious. A new edge detection algorithm of coal gangue image based on rough set theory is put forward and this algorithm is used to detect the edge of coal gauge image. At the same time, this new algorithm is compared with other tradition edge detection methods.Usually, the recognition of coal gangue image based on the mean gray and variance deviation property is used in the tradition pattern recognition algorithm. According to the characteristic of coal gangue image itself,a new method to distinguish coal gangue image based on rough set theory is put forward in this paper. The texture characteristic of coal gangue image is also considered in this method. Firstly according to different images of coal and gangue, the gray histogram and the gray level grows matrix (GLGCM) are used to extract gray level and the texture information,and then fifteen characteristic values are calculated respectively and simulated by Matlab7.0. The decision table of coal gangue image is built and discretized from characteristic vector. Next the decision table attribute is reduced on the basis of rough set theory, and then decision rule is gained. Finally according to the regulation confidence size, Rose2 software regulation matching is used to realize image recognition. From the detection result, the detection accuracy of coal gangue image based on rough set theory recognition algorithm is 87% approximately. Because coal gangue is mixed up with dust and other things in industrial environment, the identification accuracy has some deviation.