The Application Of High Resolution Image In The Identificati On Of Typical Objects In Rare Earth Mining Area

Using high resolution remote sensing images can quickly,timely and dynamically to make a observation of rare earth mining areas,and provide technical support for the sustainable development and ecological protection of rare earth mines.The existed research most concern was using the remote sensing images to extract the city's buildings,roads and vegetation and other objects.For complex features of rare earth,this paper presents a method which is using high-resolution images in rare earth typical objects.Combined with the construction of the typical features of the mining area knowledge base,to achieve the typical mining area objects recognition.The main research work and result is as follows:?The knowledge base of constructing typical feature.The interpretation signs of the objects would be established in the rare earth mining area through the analysis of diversified characteristics of typical objects of mining area.According to the interpretation signs and the Opencv graphics processing library,to be obtained the histogram features of each typical features.The AdaBoost method is used to realize the construction of typical knowledge base of rare earth mining area by trained the samples.?Selecting the optimal segmentation scale.Based on the segmentation of multi-scale algorithm,a method of optimal segmentation based on knowledge base is proposed.The histogram features are calculated for the target objects after segmentation.Then,combined with the established typical features knowledge base to select the most suitable segmentation scale.The knowledge base could be expanded by the typical features of the optimal segmentation scale.?Identification of typical objects of rare earth mining area would select pieces of non homologous high resolution image as data source and regard Gannan rare earth mining area as the study area.The object features of the study area was determined through field investigation and auxiliary data.Then,image preprocess would be carried out for two high-resolution images respectively.Secondly,the typical objects of image interpretation signs and features of knowledge base was established and the image was segmented through using the multi-scale segmentation algorithm of region merging.Besides,the optimal segmentation scale of each feature would be choosed by the method of the optimal segmentation scale based on knowledge base.Finally,the typical objects of the optimal segmentation scale combined with the features of knowledge base for recognition.?Precision evaluation means selecting user accuracy,production accuracy,overall accuracy and Kappa coefficient to analyze the recognition results.For the results part,the Kappa coefficients of the two images respectively reached 78.75% and 79.81% and the recognition effect is perfect.Therefore,experimental results show the feasibility and the universality of the method for the identification of typical objects of rare earth mining area.