The Study Of The Differencing Change Detection Method Based On Textural Features

Image change detection techniques aim to detect the changes in the same area over time. They are widely used in many fields such as environment monitoring,land use monitoring,crop monitoring,deforestation assessment and damage assessment. They can also be used to detect the urban land use and development. Synthetic Aperture Radar(SAR) is good information resource for change detection because of its all-time/all-weather capability. This paper studies urban change detection with SAR images.This paper discusses widely-used change detection techniques and mainly studies image differencing method. Due to the abundant and steady textural features on urban environment in SAR images, a texture differencing method that substitute textural features for gray information is presented in this paper. The multi-feature differencing method is also presented on the base of single feature detection and an improvement on detect performance is obtained by utilizing the complementarity among features. The Bayes rule for minimum error and supervised parameter estimation for Gaussian mixture are used to solve the problem of threshold selection and get a good experimental result. The expectation-maximization(EM) algorithm and its use conditions are also discussed to automatically get change threshold. In order to evaluate the change result, a original land cover data is defined through field survey and some evaluate standards suitable for change detection are used. Some features and feature groups that can well show the characteristics of urban environment are founded through experiment to compare the performances of different features.For all the researches in this paper, the Radarsat images in 1998 and 2000 and the Spot image in 2000 of Beijing and field survey data are used to validate. The experimental result demonstrates that a higher change detect accuracy can be obtained through the texture differencing method than through the image differencing method based on gray level and the wide application of texture differencing method for urban change detection in SAR images.