基于背景分类的弱小目标检测算法

摘要

为了降低复杂背景对弱小目标探测的影响,本文提出了一种基于背景分类的弱小目标检测算法.背景分类的依据是背景的复杂度,本文用熵运算估计背景的复杂程度,同时为了简化算法,采用多分辨率的小波变换.再用模糊数学中的隶属度函数映射到模糊特征平面,作为背景因子图像.接着对背景因子图像进行最大熵阈值分割,将复杂的地面背景与平坦的天空背景区分开来,实现了背景分类.再在罗宾逊滤波预处理后的图像基础上,争对两类背景的特性分别进行阈值分割,提取目标,从而削弱复杂背景对弱小目标探测的影响.最后实验结果表明该算法可以降低复杂背景时弱小目标的干扰,提高被复杂背景淹没的弱小目标的探测率.%A dim-small target detection algorithm based on the classification of the background is presented to reduce the influence of the complexity of the background. The classification of the background is based on the complexity of the background. The complexity of background is estimated by the entropy of the image. And the lower-resolution wavelet transform idea is introduced in order to simplify the algorithm. Then the entropy of the image is mapped to the fuzzy plane as a background factor image by the fuzzy membership function. The complex background is separated from the flat sky background to realize the background classification through thresholding the background factor image with the maximum entropy threshold segmentation. Next the target is separately segmented by using the Robinson filtered image based on the two type of background characteristics to reduce the influence of the complexity of the background. The experiments show that this algorithm can reduce the interference of complexity of background for small objects, and it can improve the probability of the point target detection.