The Comparison Of Land Cover Change Information Extraction Methods Pixel-based

Global change will affect human survival environment in the long-term, so it has gotten the extensive attention in the international scientific community in recent decades. Land cover change is an important component in the global change research, and it includes three aspects of contents:land cover change detection, driving force analysis, updating and mapping. Among them, land cover change detection is the premise and foundation. In this paper, image differencing, image ratioing, NDVI differencing and PCA differencing technique based on pixel are used for land cover change information extraction experiment. The main work is as follows.(1) Comparison between single change detection methods:Four methods which are mentioned in this paper are used to do change information extraction experiment for land cover of Shanxi province, and their extraction accuracy is evaluated. The results show that PCA differencing own highest change detection accuracy when four methods are used for land cover change information extraction experiment. But overall speaking, the accuracy of various single method is still low (user accuracy less than 80%).(2) In order to obtain better change information extraction effect, the sameness and dissimilarity of extraction results of four methods are analyzed in the basis of single change information extraction method. The results show that:differencing and PCA differencing have close extraction accuracy (the same change pixels reached 76.38%). At the same time, the extraction accuracy of NDVI differencing has greater difference compare with the other three methods. This situation is mainly due to a variety of methods has different degree of sensitivity for different surface cover. NDVI difference method is more sensitive to change of vegetation cover, so its change detection effect for vegetation is better. At the same time, NDVI differencing is relatively slow to change of water, so its change detection for water is ineffective.(3) In order to improve the change information extraction accuracy, combination of four methods to be applied according to evaluation results of the analysis of the sameness and dissimilarity. There are four different combination forms:1) The union of ratioing and PCA differencing; 2) The union of NDVI differencing and PCA differencing; 3) The union of NDVI differencing, ratioing and PCA differencing; 4) Firstly, calculate the intersection of differencing, ratioing and PCA differencing, and then calculate the union of the result and NDVI differencing. The results show that:the detection results which are generated by the union of image radioing, NDVI differencing and PCA differencing contain all the change pixels detected by four simplex methods. At the same time, this combination form reduces the over-detection phenomenon. So, the article argues that it is the best combination form.