Research Of Change Detection Based On Levene Test And Fuzzy Theory

Change detection is an important technique of remote sensing information processing and application.Sensors obtain the same area and the multi-temporal remote sensing images,through collaborative processing and analysis of men and computers,different land-use information between images can be extracted in time,change detection plays a crucial role on national census geography,prevention and evaluation of disaster,land cover / utilization detection,etc.Because the high resolution remote sensing images provide statuesque,richly detailed and brightly colored land-use information,so the use of high-resolution remote sensing images in change detection has become one of the most popular researches.According to the characteristics of high resolution remote sensing images,this paper maked full use of its spectral and texture details,used the single image spot as the basic analysis unit,presented a method of change detection based on Levene test and fuzzy set theory,and the experimental results showed that the proposed methods can effectively improve the accuracy of the change detection.The main work of this paper includes the following points:(1)Levene test was used for rough extraction of changed information based on spectral features,the relative radiation differences were alleviated between the spots for comparison by transforming the original data,the spectral differences were described between the spots of between-class and within-class by constructing difference vectors,and the amount of original data was greatly reduced by means of significant difference analysis in the premise of relatively low omission.(2)Spectrum,texture and hue information extracted from the high resolution remote sensing images were used to construct multiple feature vectors,change intensity of the spots were counted by change vector analysis,several little spots were selected as training samples after a small-scale segmentation,the best threshold was determined by the means of two iteration based on training samples.(3)Based on fuzzy set theory,fuzzy mappings from multiple feature sets to change membership were implemented by two membership functions.With independent variable increases,the membership was also increasing.According to convergent characteristics of two membership functions,the final membership was determined by the corresponding rules,thus the changed terrain information was extracted.