一种有效降低虚警率的变化检测算法

研究了图像变化检测算法,针对传统的直接比较算法对光照、噪声等变化因素敏感,虚警率较高的问题,提出一种新的变化检测算法加以解决.首先,采用提出的基于二维直方图的最大熵方法对参考图进行分类,然后利用非线性估计方法,对参考图进行估计,并利用分类结果对其进行标注,最后,通过计算测试图像像素点到估计图像对应像素点所在类的马氏距离,来度量测试图像的变化情况.通过实验比较,验证了算法的有效性.

An Efficient Change Detection Method for Decreasing False Alarm

This paper research the change detection algorithms.In order to reduce the high false rate in direct comparison methods,a new approach was presented.First,we use a method based on 2D histogram and maximum entropy criterion to classify the reference image,Second using a nonlinear method to estimated the reference image,and labeling the result by the classification result,finally we use mahalanobis distance to measure the change.At last,some experiments were provided to prove the effectiveness of the method.