共查询到15条相似文献,搜索用时 62 毫秒
1.
2.
3.
4.
5.
6.
7.
8.
介绍了自制的具有超短光路和折迭式结构的集成化学相关器。采用蒙特卡罗迭代算法对匹配滤波器进行优化设计,实验证明,该方法提高了模式识别的准确度。 相似文献
9.
10.
11.
N. Demoli 《Optics & Laser Technology》1996,28(3):215-221
A measure describing an overall estimate of the feasibility of correlation filters has been previously reported. The feasibility estimate (FE) of correlation filters is now tested in the presence of an additive stationary non-zero-mean noise. A simple noise model is used to describe quantitatively the white noise case as well as the increase or decrease of the input scene power. The usefulness of the FE is demonstrated by presenting the results of computer simulations on the example of quasi phase-only filters. The decision curves are given for the confidence interval of 0.95 probability of finding the true-class correlation peak values as a function of the total input signal power. 相似文献
12.
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。 相似文献
13.
14.
N. Demoli 《Optics & Laser Technology》1994,26(2)
The dependence of the output correlation signal on the filter parameters, filter types, and signal types in matched filtering is discussed. On the basis of the established filter criteria—the signal-to-noise ratio, peak-to-sidelobe ratio, and diffraction efficiency—a new measure is introduced giving an overall estimate of the feasibility of filters. According to that single measure, various filter types are compared and selected by numerical experiments. 相似文献
15.
A new technique called indexed composite filters (ICFs) for multiple object pattern recognition is proposed. Contrary to the
usual so-called composite filters, the proposed ICF offers the possibility to identify which object of the learning class
is detected at the output. The major advantage of ICF is that the output gives for every object detected two correlation peaks
encoding the position and identity of the detected object. Computer simulation results are given using alphabetic input scenes
that illustrates the efficiency of the proposed method. 相似文献