论文部分内容阅读
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 the photoelectric hybrid matching filter correlation recognition system, the accurate alignment of the corresponding frequency components can not be directly achieved because the input image spectrum and the matched filter are formed differently. This problem needs to be solved by scaling the matched filter. According to the Fourier transform property of the lens and the computational holographic theory, the expression of the spectrum of the input image and the expression of the transmission function of the filtered spatial light modulator after the matched filter is loaded are deduced respectively. Based on these two expressions, The formula that matches the scaling of the filter is then validated in the actual optical recognition system for the scaling calculated from this formula. Experimental results show that the quality of the correlation peak is greatly improved at this scaling ratio, which shows that the frequency spectrum of the input image is well aligned with the corresponding frequency component of the matched filter.