双Level Set遥感影像建筑物检测

Building detection in remote sensing image based on double level set function

对传统的基于梯度的Level set方法和基于区域的Chan-Vese分割模型进行分析,提出结合局部梯度和同质区域全局均值信息的双水平集遥感影像分割模型,并利用变分法得到曲线的演化方程。该模型引入一种内部约束能量项近似地表示符号距离函数,使算法摆脱了重新初始化符号距离函数的缺陷,提高了曲线的演化速度,且可以同时分割出多类目标。在建筑物检测中可以同时检测出建筑物的向阳区域、屋顶和阴影区域。实验表明,该方法是一种有效的建筑物检测方法。

By analyzing traditional level set methods based on gradient and Chan-Vese model based on regions,a double level set model for remote sensing image segmentation is proposed.This method combines local gradients and global regions average values,and uses variational methods to get the associated curve evolution equation.An internal energy term is introduced to force the level set function to be closed to a signed distance function,which can completely eliminate the need of the cost-ly re-initialization procedure.Thereforei,t can speed up the curve evolution.At the same timet,he model can segment more than two targets simultaneity.In remote sensing imaget,his method can detect the area with a sunny exposures,hadow region and the roof of the building at one time.The experimental results show that it is an effective approach to building detection.