A surface water body extraction method based on domestic remote sensing imagery of high resolution

Due to the high spectral similarity existing in water and shadow, extraction of remote sensing imagery is easily confused and misclassified. To address this problem, we propose a method combined with the object-oriented image segmentation and the artificial bee colony algorithm (ABC) to extract surface water body from remote sensing imagery. Firstly, a series of statistic factors, such as spectrum, ratio and sharp features, are calculated during image segmentation. We used these factors to make up the defect of insufficient information existing in high-resolution imagery. Then, with the strength of solving complicate problem by ABC algorithms, we chose the geometric mean of accuracies between surface water bodies and shadows as the fitness function of classifier to generate the optimal extraction rules. The experiments are carried out in the Dadeng island of Xiamen in Fujian and part of the city of Zixing in Hunan, which are based on the domestic GF-1 and GF-2 remote sensing imageries. The results are compared with the SVM classifier and show that the proposed method can achieve better overall accuracy and Kappa coefficient, indicating that the proposed method is suitable for extraction of surface water bodies from remote sensing imagery of high spatial resolution.