Coal Rock Identification Model Integrating Improved CLBP And Sensory Wilderness Theory

Automatic coal rock identification technology is one of the key core technologies for the intelligent and unmanned underground mining workings.Analysis of the characteristics of coal rock images shows that the information of the concave and convex areas in coal rock images is very important.The existing image feature extraction algorithm is not effective in describing the high-order feature information of coal rock images.To address the above situation,this paper proposes a coal rock feature extraction algorithm with improved high-order differential median-complete local binary pattern(CLBP),and explores the feature information dimensionality reduction strategy,based on which a coal rock recognition model incorporating improved CLBP and perceptual field theory is constructed.The main work is as follows.(1)To address the problem that the original CLBP algorithm only describes the local gradient features of the image,ignoring the higher-order feature information of the more important concave and convex regions in the coal rock image,while the noise immunity of the algorithm is weak.In this paper,we propose a higher-order differential median feature extraction algorithm,introduce a median processing method to strengthen the noise immunity of the algorithm,and construct higher-order differential median CLBP?S and higher-order differential median CLBP?M descriptors,which can realize the extraction of higher-order feature information of coal rock images.The algorithm fuses the higher-order differential median CLBP?S and CLBP?M descriptors with the original CLBP descriptors to reduce feature redundancy while ensuring the complete richness of feature information.(2)The coal rock identification model is constructed by fusing the improved CLBP with the perceptual field theory.This paper introduces the method of perceptual wilderness theory to carry out dimensionality reduction,and compares the performance difference between the coal rock recognition model fused with improved CLBP and SVM and the coal rock recognition model fused with improved CLBP and perceptual wilderness theory.(3)Based on the theoretical basis of the above coal rock identification model,experimental studies on the performance of the identification model are conducted.The study shows that,compared with the original feature extraction algorithm,the accuracy of the proposed fusion original CLBP and higher-order differential median CLBP algorithm is improved,and the fluctuation of accuracy with noise intensity is smaller,and the improved CLBP algorithm has better robustness.At the same time,the recognition model based on the perceptual field theory can still maintain an accuracy of over 98.5% with a 97% saving in time and space.