The Coal-rock Interface Recognition Research Based On The Digital Image Processing And Clustering Technology

China is a country with rich coal, coal resources has been an important resource in our country’s development.Coal has been and will be more than 50% in China’s energy structure for a long time. With the development of coal mining technology of coal mine working face in the mine, most of the working face has implemented comprehensive mechanical drum shearer supporting hydraulic support of coal mining method.With the continuous development of coal mine automation and control technology, the coal industry has emerged in a variety of high-tech applied in the coal mine working face. More and more mine working face has achieved high level of automation at home and abroad. The ultimate goal of the researchers in the coal mine industry is to realize the unmanned mining which is mining face area unmanned in the coal mining process.As a key technology in the unmanned mining, coal rock interface recognition problem has been studied many years by many experts and scholars at home and abroad. But the problem still has not been solved in the actual mining process. Coal rock interface recognition has become the technical bottleneck of the unmanned mining.The main object of the study is the coal and rock interface recognition problem in the process of coal shearer mining on the coal working face. Many image processing technology are used to solve coal rock interface recognition, such as histogram equalization. The paper firstly describes the environment of mining working face, and uses some improved image processing technology to optimize the images of the mining working face. And then, according to the characteristics of coal and rock, according to the characteristics of coal and rock, the paper has analyzed the feasibility of the coal-rock interface recognition solution by some basic method of image processing applications, and confirmed that the recognition can be solved by the digital image processing.According to the arrangement of the equipments on the mining working face and the control methods of the shearer drum, the paper presents a control model which can be used to solve coal rock recognition by image processing when the shearer drum working. The study confirms the parameters of the image acquisition device by the information of the hydraulic supports and shearer. The paper also describes the workflow of the whole recognition system, and expounds the fuctions of the major devices’ and how to work with each other.In the research of coal rock interface by the image of the mining face, we analyze the images by the K-means algorithm which is based on the basic theory of clustering, and find that the classification may be effected by the background imformation which is not useful for our research. The partical analyzes the veins and feature of the coal rock image by the GLCM which proves that the background imformation brings bad effect to the research. We also receive the mainly feature of the coal and the rock on the research mining by the GLCM. And then, the research reduces the background imformation by the regional growth which weaks the negative impact on the classification and identification. The paper extracts the grayscale feature of the “coal” sample and the “rock” sample which can be used as cluser centre grayscale and improves the images’ grayscale by the special filter which is designed by the situation of the coal rock images. At last, to complete classification, the paper combines the Gaussian mixture clustering and support vector machines which are improved by supervison methods and grayscale transformation.Besides completing the classification of the coal and the rock, we need to integrate the classification imformation to get the whole coal-rock dividing line which may be used in the height control of the shearer drum. The research raises a special kind of image registration method which combines with the feature points of coal-rock interface recognition by the image registration theory. This method makes the registration coordinates more accurately which is very useful to coal-rock interface recognition system.After the realization of coal rock interface recognition and the coal rock interface recognition integration, the research establishes a simple model of mining height control drum according to the data before. The model’s main idea is that quantify the length of the images by the width of the hydraulic supports. The quantitative data can be used in the shearer control more easily. This section of the research mainly contains two parts. One is quantifying images’ length and width, and the other is establishing simple model. For the quantitative, the paper raises the calculation method of the quantifying and analyzes its accuracy. And for the simple model of mining height control drum, the paper suggests the coal rock interface of the first hydraulic support as zero, the horizontal direction as abscissa which is used for location imformation and the vertical direction as ordinate which is used for mining height imformation. So, we complete establishing the simple model which can be used to direct shearer drum height control.At last, the paper analyzes the images which are collected by the image acquisition equipment on the experimental working face by the application of the partical. The result brings positive guidance to the research on coal-rock interface recognition.The main innovation of this paper includes:(1) The research combines the GMM, SVM and the improved special filter to solve coal rock recognition. The partical raises a kind of spatial filter which can make the coal and the rock difference greater.The partical raises a kind of coal-rock interface recognition method which is based on semi-supervised Gaussian mixture clustering. Normally, Gaussian mixture clustering is unsupervised, but in this research we extract the grayscale feature of the “coal” sample and the “rock” sample to guide coal-rock classification which can lower the error rate of the coal-rock classification. The partical uses a kind of support vector machine which is improved by the spatial filter to research coal-rock recognition. Some scholars has studied coal-rock classification by support vector machine based on texture feature, and this research is based on the unique coal-rock gray scale change on the mining working face.(2) According to the arrangement of the equipments on the mining working face and the control methods of the shearer drum, the paper presents a control model which can be used to solve coal rock recognition by image processing when the shearer drum working.(3) The paper raises a special kind of image registration method which combines with the feature points of coal-rock interface recognition by the image registration theory. And the research establishes a simple model of mining height control drum according to the analysis of data.In future research, there are still many points need to be improved. First, though the study has improved the versatility of the coal and rock recognition algorithm, the universal of the algorithm is still insufficient. In future, we should get more data and images from different mining working face and improve the versatility of the coal and rock recognition algorithm. Second, accumulative error may get bigger with the integration of coal rock interface recognition. In future, we can improve the integration with the help of the variety sensors’ data and the hydraulic support’s relationship with the image acquisition device. In addition, the mining height control drum model is only a simple two-dimensional model which can be improved to three-dimensional simulation mining face by the introduce of the gyro and the variety sensors.