SVM Classification with Multi-texture Data of ZY-102C HR Image

Based on a ZY-102C HR Image, three kinds of textures, the variogram texture, gray level co-occurrence matrix texture, and gradient texture, were extracted. Then, we present a new SVM classification method with multi-source data by integrating the spectral information and these three different textures. The classification result was compared with results using Maximum Likelihood Classification(MLC) and Decision Tree(DT) method. The study shows that:(1) Variogram texture and gradient texture involved in multi-source data can effectively improve image classification precision with an overall accuracy from 85.14% to 87.43% and Kappa coefficient from 0.82 to 0.85;(2) The variation function of absolute value form provides a theoretical basis for the optimal texture window analysis, and textural features based on the average step can significantly improve classification accuracy with an overall accuracy from 75.2% to 87.14% and Kappa coefficient from 0.7 to 0.87;(3) Based on multi-source data, the SVM classification method for high spatial resolution remote sensing images can effectively overcome the fragmentation problems associated with traditional image classification methods. Our results were significantly superior to MLC and DT with an overall accuracy of 89.14% and Kappa coefficient of 0.87;(4) The resource satellite data ZY-10C has a certain stability and advantages for the extraction of winter wheat.