| In this study, we chose north-central Shawan County in Xinjiang Uygur Autonomous Region as the study area. Multi-temporal remote sensing images have been widely used in cotton extraction, but these studies mainly focused on sole feature such as NDVI. An effective method of integrated multi-features based on multi-temporal images was proposed to extract cotton information. Nine scenes of Landsat8 in 2013 were collected, NDVI time series, optimal temporal reflectance image and texture features were combined as the original classification features. When the training samples were sufficient, if the feature was excessive, the classification accuracy may decrease due to the redundant information. In this paper, we studied the feature selection algorithms which were suitable for several types of features, and then the optimal features were used for classification and cotton extraction. Finally, we evaluated the accuracy of classification results by confusion matrix. The results as follows:(1) Feature selection algorithm based on improved_OIF was proposed. The method focused on the information content and correlation of the feature combination,the separablity between classes also was taken into account. Using the optimized set based on improved_OIF algorithm for classification, the result was superior to the classification result based on the optimized set of OIF algorithm. It showed that this algorithm was effective to improve the OIF algorithm performance in the feature selection;(2) Compared with the classification results of NDVI time series, the method of integrated multi-features based on multi-temporal images can effectively improve the classification accuracy;(3) Compared with the classification results of the original features, the optimized set based on the improved_OIF algorithm and the optimized set based on the rough set algorithm achived higher accuracy. It showed that these two feature selection algorithms not only reduced the classification complexity effectively, but also improved the classification accuracy. The effectiveness of improved_OIF algorithm and rough set algorithm in multi_feature selection was verified. |
