研究期間：10108~10207;Land Change Detection Using Multi-temporal Satellite Images by Applying Pseudo Invariant and Variant FeaturesBecause of the incessant improvement of sensor’s technology, the temporal and spatial resolutions of acquired satellite images are keeping increasing. It means that it is more and more difficult to obtain change detection results using manual image interpretation. Therefore, the main objective of this project is to develop an automatic land change detection model using multi-temporal satellite images. For this purpose, the normalization for images of different dates and the selection of thresholds for detecting changes are essential works need to consider. This project is designed for the duration of three years. In the first year, the Pseudo Invariant Feature (PIF) will be extracted from the source images and used to generate a set of normalized multi-temporal satellite images for change detection. Then, the concept of PIF will be extended to develop the Pseudo Variant Feature (PVF) which can be applied to develop the change detection algorithm using multi-temporal satellite images. In the second year, the major work will be focused on the research about the method of automatic change detection. In order to improve the drawbacks caused by pixel based change detection approaches, an object based change detection model will also be developed according to the image object generated by image segmentation technique. Then, based on the PVF obtained in the first year, a statistics test approach will be developed to detect the images changes. In the third year, a semi-supervised change detection approach will be proposed. First, the PIF and PVF provided by the first two years will be used as a training sample to initialize a Support Vector Machine (SVM). Then, by using the decision boundaries provided by SVM, a more stable and accurate change detection result is expected to be achieved. In addition, the performance of the method developed in the first two years will also be compared with the SVM approach. This project mainly can provide a better solution for change detection using multi-date and multi-source satellite images. In the feature, the detected change results by this project can be used in related land management applications
