Research On Automatic Classification Methods Of Feature Information Based On Panchromatic Remote Sensing Images

With the rapid development of Chinese economy,urbanization development policy is being carried out in full swing,so the demand for land resources is also rising.Using the remote sensing image to classify and extract the typical urban features has become the mainstream trend.It is one of the most important research topics to use the suitable algorithm to extract the information of the typical urban features.In this paper,Jilin-1 panchromatic A-satellite panchromatic remote sensing images are used to classify and extract four types of buildings,roads,woodlands and meadows in the region using pixel-based and object-oriented classification methods.As follows:(1)For the pixel-based classification method,The paper mainly researches the supervised classification which includes minimum distance method,maximum likelihood method,BP neural network method,and support vector product method,and at the same time studies the unsupervised classification which includes ISODATA algorithm and K-means algorithm.(2)For the object-oriented classification method,based on the spectral,shape and texture features of the image,Sobel edge operator segmentation and full Lambda-Schedule segmentation are used to segment the image,and then use the fuzzy classification method to establish information extraction rules that analyze the characteristics of each object.(3)Research the classification accuracy evaluation method that use confusion matrix.According to the classification results of each algorithm,the overall classification accuracy and Kappa coefficient are used to evaluate the overall classification accuracy.For the classification accuracy of single-class objects,that used three types of indicators to be evaluated,misclassification error,omission errors and single classification success.The experimental results show that for the pixel-based classification method,the classification method with the highest overall classification accuracy is the maximum likelihood method in supervised classification,the overall classification accuracy of 83.8680%,Kappa coefficient of 0.7561.The single-class classification of roads,buildings and meadows was the most accurate,and support vector product method has the highest classification accuracy when single-class extraction of forest land.For the object-oriented classification method,the overall classification accuracy is 94.4721%,Kappa coefficient is 0.903.On the whole,the classification accuracy of object-oriented method is higher than that of pixel-based classification.The results of this paper are of great significance to the development of urbanization in Changchun,and also serve as a guide for the classification of Jilin-1 satellite images.