Research On Segmentation And Classification Of Object-Oriented High-resolution Remote Sensing Images

In recent years,industrial solid waste pollution has become one of the top ten environmental pollution problems.With the continuous development of industrial production,the number of industrial solid waste is increasing.The problem of industrial solid waste exists: First,the impact of urban and other beautiful;Second,the industrial solid waste in the harmful ingredients will be with the rain into the soil caused by water and other pollution,resulting in secondary hazards,The understanding of the distribution of industrial solid waste can be a good grasp of the pollution situation.The accuracy of segmentation results of industrial solid waste remote sensing image directly affects the judgment of its monitoring results.However,the traditional image classification is based on pixels,does not take into account the characteristics of high-resolution image information,which is not very high classification accuracy.In view of the above problems,object-oriented thinking is applied to the segmentation and classification of images,that is,from the traditional pixel-based processing into object-based processing;Through the full use of shape and spectral characteristics,so that the polygon of the object is closer to the boundary of the real object.The main ways of segmentation and classification of remote sensing images of object-oriented industrial solid waste in this paper can be summarized as follows:(1)Before the remote sensing image processing PCA processing segmentation,and then quadruple the tree segmentation of the image processing into a certain image block,that is,image objects;(2)Calculate the weight between blocks and blocks based on the criteria of the least heterogeneity;(3)The use of a rapid merger method will be divided into blocks and blocks between the largest heterogeneity,the smallest homogeneity of the object,that is,the final segmentation results;(4)The SVM method of active learning is classified according to the obtained segmentation result,and the final classification result is compared with some traditional methods.In this paper,the results of segmentation,classification and traditional methods and some corresponding improvement methods are compared and analyzed.The results of classification are homogeneity,heterogeneity and PRI.The classification results are relatively good in terms of overall accuracy and kappa coefficient effect.