| Summary: | 碩士 === 元智大學 === 電機工程學系 === 90 === Abstract
In our previous published papers on the non-zero order opto-electronic joint transform correlator, classical training images are put around the center to synthesize the reference function. However, it can’t guarantee that minimum average cross correlation energy can be achieved.
To improve this, we shift training images pixel by pixel from the left to the right and from the top to the bottom. We record the power spectrum and calculate the reference function to find minimum average cross correlation energy while shifting the training image each time. By this procedure, the optimal training image database will be found. Compared with the original method, our method will get sharp correlation peaks and less sidelobes, which can increase the recognition efficiency and reduce false alarm rates.
|
|---|
