Research On Detection And Recognition Of Coal And Rock Feature Information Based On Spectral Imaging Technology

Coal is still the main energy in the future.With the improvement of intelligent degree of fully mechanized mining face,the lack of tunneling efficiency leads to the imbalance of mining in many coal mines.The identification of coal and rock in the process of roadway excavation is the premise to realize the adaptive cutting of roadheader,and it is the key to realize the intelligent and unmanned roadway excavation.The low illumination and high dust of heading face seriously limit the accuracy of coal and rock identification,and pose a challenge to adaptive cutting,which has become a key problem for the intelligent development of heading face.Therefore,this paper puts forward the detection and recognition method of coal and rock feature information based on spectral imaging technology.Combined with the image feature information collected by multispectral camera and spectral band feature information,the coal and rock recognition model is established to realize the detection and recognition of coal and rock feature information under harsh working conditions,solve the problem of low recognition rate of coal and rock in harsh working conditions,and provide strong support for the intelligent,less humanized and even unmanned efficient tunneling of tunneling face.Aiming at the problem that it is difficult to distinguish the band characteristics and image characteristics of coal and rock,the image information under multi-band is selected by using the image dimension reduction method,and the preliminarily determined image information is filtered by using the filtering algorithm.The texture information of the preprocessed image is collected by using the gray level co-occurrence matrix,and the four texture characteristic parameters of energy,entropy,contrast and homogeneity are finally determined as the coal and rock image information.The competitive adaptive reweighted sampling(CARS)and successive projections algorithm(SPA)and random frog(RF)and uninformation variable elimination(UVE)based on spectral feature information are studied.Finally,four single spectral features are obtained as coal spectral information.Aiming at the redundancy problem of data after simple combination of coal and rock image information and spectral information,the multi-information data fusion technology of coal and rock is adopted,and the dimension reduction fusion method of feature layer based on principal component analysis(PCA)and linear discriminant analysis(LDA)method is proposed.The dimension reduction fusion of all spectral band information and texture information,and the feature information after simple combination of spectral band information and texture information selected by four band extraction methods are carried out.The feature information after dimension reduction fusion of coal and rock is determined by analyzing the results of information dimension reduction.Aiming at the problems of few coal samples and large similarity,the modeling method based on least squares support vector machine(LSSVM)and logistic regression(LR)is studied.The modeling process based on coal and rock feature information recognition is proposed.The single feature information,simple combination feature information and dimension reduction fusion feature information of coal and rock are modeled respectively.The best coal and rock feature information detection method and recognition model are determined by analyzing the classification view of modeling results.Finally,the coal and rock feature information detection and recognition platform is built,and the software system suitable for coal and rock feature information detection and recognition is developed.The functions of data acquisition,image processing,spectral band extraction,information fusion,modeling analysis and coal and rock recognition results are realized.The experimental results in good laboratory environment and simulated harsh environment of coal mine(insufficient exposure+dust)show that the detection and recognition method of coal and rock feature information proposed in this paper improves the recognition rate of coal and rock samples.At the same time,the obtained coal and rock interface distribution information is consistent with the actual coal and rock distribution,which is helpful to solve the problem of adaptive cutting in the process of roadway excavation and ensure the intelligent and unmanned roadway excavation.