Extract Reservoir Water Area Automatically Using Masked Active Contour Model from Multispectral Remote Sensing Images

Continuous monitoring on Miyun Reservoir monthly,or with an even shorter period,plays an important role on knowing the change of Miyun Reservoir water area timely.Because the monitoring period is very short,the change between two neighboring monitoring status is small and it is not easy to find the change trend due to the existing of errors while water area extracting.To reduce the system errors and human errors as much as possible,we try to extract water area from multiple Beijing-1 small satellite multispectral images using an improved active contour model,which we named Masked Active Contour Model(MACM).The MACM we proposed is based on Ron Kimme's active contour model.We improve Ron Kimme's model by extending its applying range from grey images to color images or multispectral images,introducing a mask image which automatically change during the evolution of the level set function to make the image meet the demand of bimodal image,and replacing the fixed time step with a converging one for fast convergence,etc..These improvements make the model more practical.Multiple Miyun Reservoir Water areas were extracted using the MACM from the Beijing-1 small satellite 32-meter multispectral images received ranging from 2006 to 2008.To evaluate the accuracy,Miyun Reservoir Water areas of six different periods were extracted interactively from the high resolution panchromatic images of Beijing-1 small satellite,which were received very near or even the same as the day on which the corresponding multispectral images were received.The accuracy assessment results show that the mean accuracy of the water areas extracted using MACM is 97.54% and the square of the correlation coefficient with the reference data is 0.896 1.Our researches show that the water areas produced by the MACM are accurate and stable enough compared with the reference data and comparable between different periods.So extracting reservoir water from remote sensing images using the MACM can improve the stability of continuous monitoring and reflect the change trend of reservoir water area accurately.Our research gives a new approach for the automatic extraction of water area from remote sensing images.