在光电混合匹配滤波相关识别系统中, 由于输入图像的频谱与匹配滤波器的形成方式不同, 不能直接实现对应频率成份的准确对准。这个问题需要通过对匹配滤波器进行缩放来解决。根据透镜的傅里叶变换性质和计算全息理论, 分别推导出输入图像的频谱的表达式和滤波空间光调制器加载了匹配滤波器后的透射函数的表达式, 并根据这两个表达式得到匹配滤波器的缩放比例的公式, 然后在实际的光学识别系统中对根据该公式计算的缩放比例进行了验证。实验结果表明, 在该缩放比例下, 相关峰的质量有较大改善, 这说明此时输入图像的频谱与匹配滤波器对应频率成份实现了较好地对准。

In hybrid photoelectric matched filtering correlation recognition system, the corresponding frequency components of input images spectra and matched filters can not directly aim at each other, because their shaping modes are different. This problem should be solved by magnifying or minifying matched filters. According to computer-generated hologram theory and Fourier transform characteristics of lens, spectrum distribution expression of a input image and the transmittance function expression of filtering SLM that loads the matched filter of this input image are derived respectively, and the formula of the scaling factor for matched filters is acquired according to the two expressions, and then the scaling factor calculated using the formula is validated in practical optical recognition experiments. The results of experiments indicate that there is a comparatively big improvement in the quality of correlation peaks under this scaling factor, which means that under the circumstances it is realized that corresponding frequency components of input images spectra and matched filters aim well at each other.