| The field research of water information costs lots of human and material resources. The widely used water extraction method based on the remote sensing image provides a more efficient and lower cost way monitor water information. Aim at the problem that the Cloud and shadow in satellite images usually cause inaccuracy in water extraction and difficulty in time-series analysis of water resources, in this paper, we improved the “global-local†water extraction method and presented a novel water temporal change detection technique.To improve the precision of water extraction and water change detection, we modified the “global-local†water extraction method after analyzed various water index and their advantage. On one hand we chose the appropriate water index according to the background information of image, and employed a more precision segmentation method on global segmentation, then used local extraction method iteratively to approximate the local water boundary until the results were stable. On the other hand, we proposed a water change detection technique based on an information supplement strategy using the time close images. This technique solved the cloud and shadow caused data quality problem on water change detection.The experiment on Anhui section of Huaihe River watershed showed this method took full advantage of the time close images, even if the image quality was poor, to extract a more integrate result comparing to use a single HJ-1 image. The randomly calculated check points minimized the manual intervention, thus provided an accurate and effective way to monitor the time-series of water resources. In the study area 8295 check points were extracted. The results showed the water resources of the study area in raining season(July and August) were more abundant than dry season(March and April) in 2013, especially many temporary waters had been detected in the south of the watershed in raining season. The total water area increased 22.1% in raining season compared to dry season. |
PDF Full Text Request