基于边缘约束的高分辨率遥感影像分割算法研究

Research of High Spatial Resolution Remote Sensing Image Segmentation Method Based on Edge Restriction

影像分割是面向对象影像分析的基础,已经发展出许多算法。随着遥感影像空间分辨率的提高,复杂的噪声和丰富的细节信息给影像分割带来了巨大的挑战。分水岭变换能够得到连续封闭的边缘,但存在严重的过分割现象,为了解决这个问题,本文提出Canny算子边缘约束下的分水岭分割算法。首先,对遥感影像进行分水岭变换和Canny边缘检测;以Canny边缘作为参考数据,以分水岭的结果和Canny算子检测出的边缘的重合度和相邻区域的特征差异为标准对分水岭图像进行区域合并,得到最终分割结果。实验结果表明,该算法能够有效解决分水岭变换的过分割问题,准确提取出主要地物目标的边界轮廓。

Image segmentation is the basis of object-based image analysis,and many algorithms have been developed.With the improvement of spatial resolution of remote sensing images,complex noise and rich detail information bring great challenges to image segmentation.Watershed transform algorithm can get continuous closed edge,but there is serious over segmentation phenomenon.In order to overcome this problem,a watershed segmentation algorithm with Canny operator edge constraint is proposed.Firstly,watershed transform algorithm and Canny edge detection are applied to remote sensing image;Canny edge is used as reference data,and the result of watershed,the coincidence degree of edge detected by Canny operator and the characteristic difference of adjacent region are used as standard to merge the watershed image and get the final segmentation result.The experimental results show that the algorithm can effectively overcome the over segmentation problem of watershed transform and extract the boundary contour of the major object accurately.