针对SAR图像自动目标鉴别的应用,提出了一种基于遗传算法的特征选择方法.首先提取了反映目标和杂波虚警差异的八个特征,分别是:四个空间边界属性特征,一个分形维数特征和三个对比度特征.然后对由八个特征构成的特征矢量采用遗传算法进行特征选择,以选出对于目标鉴别最优的特征序列.遗传算法中适应度函数的设计综合考虑了描述长度、鉴别总错误数以及漏报数等三个因素,使得该适应度函数对于特征优劣的评价更全面.实测数据的实验结果证明了所提算法的有效性.
For the application of target discrimination in SAR images,a method of feature selection based on Genetic Algorithm (GA) is proposed in this study. First, eight features including four spatial edge property features, one fractal dimension feature and three contrast features, are extracted to distinguish target and clutter false alarm. Then, the GA is applied to select the best feature series from the feature vector consisting of eight features. The fitting function in GA is devised by considering three factors including describing length, the total wrong number of discrimination and losing number. So it can evaluate the feature better than other algorithms. The experimental results of actual data demonstrate the presented algorithm is effective.