基于主成分分析的MACH滤波器设计

Design of the MACH matched filter based on principal component analysis

针对解决畸变不变目标识别中选择目标样本数目过多的问题,提出了一种基于主成分分析的MACH匹配滤波器设计方法,这是使用去除冗余的本质特征图像来替代目标样本图像的一种方法,能够减少匹配滤波器设计的样本数量,在对一定范围内的畸变目标进行识别,只需要少数几个匹配滤波器就可以完成,而且同一匹配滤波器还可以对不同的畸变方式(如尺度缩放、角度旋转、运动模糊失真)目标进行识别,实验结果表明,这种匹配滤波器在较大的畸变范围内仍然能够得到较尖锐的相关峰。

For settling the problem of aberrance targets stylebook amount overabundance,the design of PCA-MACH filter based on the principal component analysis was proposed,the method used the essence character images that wipe off redundance to substitute the target stylebook images,it can be reduced the amount of target stylebook for the filter.When the target distortion is confined to a certainty range,only need fewness matched filters can recognize them.The same filter can be used to recognize differant aberrance fashion（such as the size room,angle aberrance,movement blur distortion）,the result of experiment indicated the very acuminate correlation peak is gained by the matched filter.