Research On Identification Algorithm Of Coal And Gangue Based On Image Feature

Coal is a kind of basic energy, accounts for about 70% in primary energy consumption structure, and quite a long period of time in the future, will till holds the first place. The gangue mixed in the coal, has low carbon content and low calorific value, not only reduces the quality of coal and reduces the combustion efficiency, but also releases harmful substances when burned that can aggravate environmental pollution. The identification of coal and gangue is the most effective way to solve the problem, and is an indispensable step in the process of production in the coal mines, therefore, the identification of coal and gangue is of important practical significance.Most of the coal mines still adopt the method of artificial identification, but the identification rate of this method is low, and the underground working condition is bad,it is not conducive to the staff’s physical and mental health. In this topic, according to deficiencies in the present situation of the identification of coal and gangue, digital image processing technology and pattern recognition theory is applied to the automatic identification of coal and gangue, and can lay a solid theoretical basis for the deep study in the future and its application in actual system.This topic in view of coal and gangue images, puts forward the method of combining gray level co-occurrence matrix and wavelet transform to extract the texture features of coal and gangue images, and uses the support vector machine that which parameters are optimized by the particle swarm optimization algorithm to identify coal and gangue images. This topic extracts the mean and variance of coal and gangue images as gray features, analyzes the principle of gray level co-occurrence matrix and wavelet transform respectively, and uses gray level co-occurrence matrix to extract the contrast, correlation, energy, deficit moment of the coal and gangue images as texture features, and uses the Symlet4 wavelet to decompose the coal and gangue images into three layers and extracts the mean, variance and energy of each sub-images as texture features, based on the analysis of the basic principle of support vector machine, the influence of kernel function and its parameters to the performance impact of identification, uses the particle swarm optimization to optimize the parameters, and through the optimized support vector machine based on the extracted gray features combined with texture features to identify the image is coal or gangue, and verifies the feasibility and validity of the algorithm on Matlab, and confirms that the proposed algorithm has higher recognition rate.