一种目标级的遥感图像变化检测算法

An Object-level Change Detection Algorithm for Remote Sensing Images

传统的像素级变化检测方法对图像的配准准确度要求较高,因而在实际运用中受到很多限制.在人造目标检测的基础上,提出了一种目标级的基于局部配准误差补偿的变化检测方法.根据遥感图像中人造目标与自然目标的纹理差异,对图像中的人造目标进行检测和分割,再对分割图像采用提出的算法进行变化检测.实验表明,与传统的像素级变化检测方法相比,本算法具有较高的检测准确度,对配准准确度的要求也有所放宽,并且可以简化变化检测前的辐射校正工作和变化检测后的像素分类的工作.

A general drawback to traditional pixel level change detection algorithms lies in their high requirements for image registration accuracy, which leads to many limitations in their practical applications. To overcome such a drawback, an object-level change detection algorithm based on local registration error compensation is proposed. The algorithm is composed of two stages. The first stage aims at detecting and segmenting man made objects according to the texture difference between man made objects and natural ones. The second stage detects changes in the segmented images. Experiments show that, compared with traditional pixePlevel change detection algorithms, the proposed algorithm has higher detection accuracy, and lower requirements for registration accuracy. Moreover, radiation adjustment before change detection and pixel classification after change detection can be simplified.