Research On Identification Method Of Coal And Gangue Based On Image

The key technology of coal rock recognition based on image was systematically studied in this paper.The paper introduces the importance of coal gangue recognition to improve automation coal production,The traditional control of the shearer of drum and pick-up gangue works are much finished by manual completion.In order to rcalize the automatic control of the shearer drum and the intelligent picking-up gangue,the paper puts forward the image recognition method based on gray-scale and texture for the fast and accurate identification of coal and gangue.According to the locations of the mining areas,this paper builds the sample sets from Han City,Shaanxi province,consisting mainly of lean coal and shale,and from Jizhong,Hebei province,consisting mainly of coking coal and sandstone,and from Geological Institute of Xi'an branch of Coal Science Academy,Shaanxi province,consisting mixture of coal and gangue.And we set up the coal and gangue image acquisition system in the laboratory.According to the noise of the real coal production,we apply the image enhancement technique based on local gray scale transformation,and the image filtering technology based an adaptive filtering to optimize the image of coal gangue,and obtain the system result.Extracting the characteristics of coal and gangue by gray scale statistical method,we find that coal and gangue have a big difference in average and peak of gray scale,age constitute the gray-scale feature vector.And extractingthe characteristics of coal and gangue by texture method,we find that they have a big difference in gray-degree co-occurrence contrast and gray-degree symbiosis entropy,and constitute the texture feature vector.We use the peak of the gray feature and the contrast of texture feature to contrast integrative feature.We choose LS-SVM as machine learning model.The model is trained by 3 groups of parameters separately,The results show that the model trained by the integrative feature outperforms others in accuracy.The feature vector of integrative feature is LS-SVM as the classification method,which achieves 100%and 100%in the identification of coal and gangue in Shaanxi and 98.6%and 100%in Hebei and 94.0%and 96.0%in Xi'an branch of Coal Science Academy,which demonstrates the effectiveness of the classifier.