Coal and rock recognition method based on low frequency component characteristics of discrete cosine transform and learning vector quantizatio

To solve problems of small application scope and low recognition accuracy rate of existing coal and rock recognition methods, discrete cosine transform (DCT) is used to process coal and rock image blocks. DCT transform coefficients of each image block are arranged in the form of “Z” shape to express vector of image blocks. There are two extraction methods of coal and rock image features: coal and rock images feature vectors are constituted by average value of each image block vector and variance of all image blocks vector, and the feature vectors are expressed through cascading image block vector by the order of DCT transform of image block. Learning vector quantization neural network is used for coal and rock recognition. Recognition accuracy of the two feature extraction methods both achieves 96.67%. The proposed coal and rock recognition method improves recognition accuracy of 3.3% than Haar wavelet method and 5.8% than Daubechies wavelet method.