基于改进SLIC算法的Landsat 8 OLI遥感影像水体提取

Water Extraction from Landsat 8 Timas New Roman OLI Remote Sensing Images Based on Improved SLIC Algorithm

洪涝灾害时常发生,基于遥感影像绘制水体分布图观察水体变化很有必要。针对简单线性迭代聚类(SLIC)算法在遥感影像水体分割时的易误分割问题,提出一种改进的新超像素分割算法(F-SLIC)。首先基于预处理后的Landsat 8 OLI遥感影像计算新水体指数WBAWI,突出水体特征;再结合线性特征增强和局部三值模式(LTP)纹理特征获得新纹理特征(E_LTP),纳入SLIC算法对WBAWI图像进行超像素分割处理;最后对水体超像素进行区域合并,得到完整水体轮廓,通过提取同一区域不同时间的水体轮廓有效观察水体变化。结果表明,与SLIC算法相比,F-SLIC算法的边界召回率更高、欠分割错误率更低,解决了小水体难分割的问题。

Floods occur frequently,and it is necessary to draw a map of water bodies distribution based on remote sensing images to observe the change of water bodies.An improved new superpixel segmentation algorithm(F⁃SLIC)was proposed to address the issue of misclassification in water segmentation of remote sensing images using the Simple Linear Iterative Clustering(SLIC)algorithm.Firstly,based on the preprocessed Landsat 8 OLI remote sensing images,the new water index WBAWI was calculated to highlight the characteristics of the water body.Combining linear feature enhancement with local ternary pattern(LTP)texture features to obtain new texture features(E_LTP),incorporating SLIC algorithm for superpixel segmentation of WBAWI images.Finally,the superpixels of the water body were merged to obtain a complete water body contour.By extracting water body contours from the same area at different times,water body changes were effectively observed.The results show that compared with SLIC algorithm,F⁃SLIC algorithm has a higher boundary recall rate and undersegmentation error rate,solving the issue of difficult segmentation of small water bodies.