Abstract: Efficient change detection method of land cover has important significance in geographical conditions monitoring. In view of the low efficiency of manual visual interpretation and the high error rate and omission rate of software automatic interpretation in current land cover detection, an automatic change detection method of land cover types based on union feature is proposed in this study. Through the comparison of seven different feature combination schemes, the optimal combination form of combination gray co-occurrence matrix, gray histogram, spectral statistical characteristics and object characteristics is established. Besides, support vector machine high-dimensional classifier is designed for classification. The experiment results show that, in the case of complex land cover distribution in Zhejiang province, the accuracy of this method is more than 85% in detecting the type change of artificial structures such as building areas and construction sites, based on the domestic high-resolution satellite images with resolution better than 1 m. Change detection of grassland and cultivated land can also achieve good results.

Key words: surface coverage, joint features, support vector machine, automatic, detect