Study On Image Recognition Technology Of Coal And Rock Identification Applications

Coal is an important raw material for metallurgy, chemical industry, it is also a veryimportant source of energy, the coal industry is an important basic industries related tonational economic lifelines and energy security. Currently, most of the coal mines in the coaland rock identification are still using manual identification method. Manual identification ofcoal with a lag, and work in the mine environment is very bad, it is not conducive to artificialobservation of coal and rock, so it becomes very important to the study of the automaticidentification of the coal-rock interface.Coal-rock interface automatic identification has a great significance to realize nobodyground remote control, it can promptly adjust the shearer drum height. Which not only canreduce the lost coal, improve coal mining rate, but also reduce the cutting rock shearer wear.This paper focuses on the study of the automatic identification system of coal and rock image,using image recognition technology to the coal mine of coal and rock automatic identification,which has laid a solid foundation for future in-depth research and practical applications.This paper studies on the image of coal and rock from a coal mine, detaileddescription of the main components of the image recognition. Through the analysis of thegray mean and the different textures of the coal and rock, this article puts forward a coal androck identification method of combining the gray mean and texture features of coal and rock,and classified with a classifier. For texture extraction, this paper researches gray levelco-occurrence matrix and wavelet transform two methods in separately. For the wavelettransform, this paper adopts the Sym4wavelet to extract texture features of coal and rock, inorder to reduce the dimension of texture features, this paper uses principal componentanalysis to reduce the dimension of texture feature, data dimensionality reduction as thetexture features of wavelet transform to extract. For classifier, BP neural network and supportvector machine (SVM) are analyzed and studied in this paper. Finally, based on thecomparison of the experimental results, it is concluded that in the case of using the sameclassifier, through the wavelet transform to extract texture is higher than the gray levelco-occurrence matrix extract texture recognition rate in general. After the gray mean iscombined with the texture features extracted by wavelet transform, the effect is better thanthat only use the texture characteristics method, and then use SVM classification has the best effect. So this paper combines gray mean with texture features extracted by wavelet transformand uses the support vector machine to classify and identify, This method has been furtherimprove the identification accuracy.