弱小目标检测算法性能评价的回归分析方法

针对弱小目标检测算法的特点,建立了一种对算法性能进行评价的机制和方法。以图像特征指标作为输入量,算法性能指标作为输出量,用响应函数模型描述二者之间的关系,通过回归分析对算法的性能做出了评价。利用所建立的方法,分别对一种基于中值滤波和一种基于形态学滤波的检测算法进行了分析和比较,结果表明,对于试验所采用的背景图像情况,后者对弱小目标的检测性能总体上要优于前者。

Evaluation method for weak and small target detection algorithm

In view of the algorithm features for weak and small target detection, an evaluation method of algorithm performance is proposed. The method is developed with response function modeling by choosing the image parameters as input and the algorithm performance as output. The statistical relation between algorithm performance and image measurements was studied through regression analysis. With the proposed method, an analysis and comparison was carried out for the performance of two algorithms based on median filter and morphological filter, respectively. The test results show that the later one would be preferred for detecting weak and small targets.