核典型相关分析的高分辨遥感影像变化检测

Change detection of high remote sensing image alteration based on kernel canonical correlation analysis

传统的线性多变量变化检测方法在处理高分辨率遥感影像变化检测时,容易出现明显的"椒盐现象"的问题。该文基于面向对象的分析思想,提出核典型相关的变化检测方法。首先对高分辨率遥感影像进行多尺度分割获得影像对象;然后运用核函数多变量典型相关分析,构造差异向量,并进行最小噪声变换,提高影像对象的信噪比;最后采用ROC曲线确定最佳的变化检测阈值。实验结果表明,该方法不仅消除了"椒盐"现象的干扰,而且提高了变化检测的精度。

The traditional linear muhivariable ＂salt and pepper phenomenon＂ when dealing with change detection method is easy to solve the problem of the change detection of high resolution remote sensing image. Based on the object-oriented analysis thought, this paper presents the nuclear typical correlation change detection method. Firstly, the high resolution remote sensing image is segmented by multi-scale segmentation. Then, using the kernel function multiple canonical correlation analysis, construct the difference vector, and carry on the minimum noise transform to improve the signal to noise ratio of the image object. Finally, the ROC curve is used to determine the best change detection threshold. The experimental results show that the method not only eliminates the ＂salt and pepper＂ phenomenon, but also improves the accuracy of the change detection.