Object-Based Cloud-Contaminated Area Detection Using Multi-temporal CBERS Images

In this paper, an automatic cloud-contaminated area detection approach is studied using multi-temporal middle-spatial satellite remote sensing images. A cloud detection method using multi-temporal images based on change detection method in remote sensing image interpretation and timing information are applied. This approach is composed of three models. Firstly, image preprocessing is applied to image data, including relative geometrical registration and radiation correction. Secondly, multi-scale segmentation is applied to multi-temporal images using commercial software Ecognition. Thirdly, difference images are obtained from multi-temporal remote sensing images using a temporal-spatial difference analysis model based on Mathematical statistics approach. And finally, accurate cloud-contaminated area detection results are obtained using difference image with threshold segmentation. Multiple experiments are carried out to test the effectiveness, robustness and applicability of this approach. Experimental results prove that this approach has the characteristics of high-accuracy and it can be applied to cloud-contaminated area detection with multi-temporal remote sensing images.