扩展等相关峰综合鉴别函数算法研究

Study on extended equal correlation peak synthetic discriminant function algorithm

为改进等相关峰综合鉴别函数(ECPSDF)算法,提出了在其约束条件中利用空域构造样本(fi(x,y)-βm(x,y))取代训练样本fi(x,y),将平均训练样本m(x,y)引入到ECPSDF算法的约束条件中,得到了扩展ECPSDF算法(EECPS-DF)。在匹配滤波器设计过程中,针对目标图像、训练样本数和训练样本畸变量间隔,选择合适的β值,使设计的匹配滤波器性能达到最佳。对飞机图像的相关识别进行了模拟和实验,结果表明,当训练样本角间距分别取6°和8°时,其η值分别由0.8584和0.6719减小为0.2145和0.1865,EECPSDF算法能较好地克服ECPSDF算法的缺点。

To improve the Equal Correlation Peak Synthetic Discriminant Function（ECPSDF） algorithm, the samples （f1（x,y）-βm（x,y）） in space domain are used to replace the training samples fi （x,y）, and the mean training sample m（x,y） is introduced into the constraint condition of ECPSDF algorithm. Thus the Extended ECPSDF（EECPSDF） algorithm is obtained. According to the target image, the number of training samples and the interval of distortion variables of training sample, the appropriate β value is chosen to optimize the matched filter. The experiment for the correlation recognition of the plane image is carried out, and the results show that the η value decrease to 0. 214 5 and 0. 186 5 from 0. 858 4 and 0. 671 9 respectively when angle interval of training sample is equal to 6° and 8°, which can improve performance of ECPSDF algorithm.