Feature Extraction Algorithm And Its Application In Coal Maceral Classification

Coal macerals refers to the basic organic component units in the coal which are identified under a microscope. It has a close relationship with the coal properties, such as CO adsorptive characteristics, cohesiveness and so on. Therefore, the implementation of the automatic classification and recognition of each maceral is beneficial for predicting the coal properties, and guiding the processing and utilization of coal. Feature extraction is an important part of the automatic classification of macerals. Extract features of appropriate number and independent each other is a key step for effective classification. In this dissertation, based on the analysis of the characteristics of the micro components in the coal microscopic image, texture features and gray distribution features are firstly extracted from the sample images of macerals; then by using feature extraction algorithm initial features are further selected; finally, a Support Vector Machine(SVM) classifier is constructed to verify the extracted features. The main contributions of this dissertation are as follows:(1) On the basis of consulting and reading related literatures, research status of maceral classification based on image analysis and feature extraction algorithms are summarized, image characteristics among different components of macerals are analyzed.(2) According to the texture and intensity characteristics of macerals, 5 texture related features as energy, entropy, moment, local smooth, maximum probability based on gray level co-occurrence matrix and 6 intensity related features as contrast, mean, standard deviation, skewness, uniformity, peak and kurtosis based on gray-level statistics are extracted. According to the distribution of different features, the effectiveness of each feature for classification is analyzed. By the classification experiments, the redundancy existed between initial feature parameters and its effect for the classification are analyzed.(3) By using Principal Component Analysis(PCA) and Linear Discriminant Analysis(LDA) respectively, primary features are further selected. Two feature extraction algorithms are analyzed and compared with the classification experiments.(4) With SVM and feature set from PCA and LDA, the classification between inertinite, vitrinite, exinite and some submacerals are implemented, correct rates of classification with different feature selecting algorithms are compared.The main innovations of this paper are as follows: The feature extraction algorithm is used to extract the features of the original feature set, which can reduce the dimension of feature space and improve the accuracy of classification.