Optical Matched Filter On The Image Recognition

Based on the basic theory of optic matched filtering correlator, the rotating invariance multiplexed matched filtering recognition method of cascade modified filter has been given with theoretical verification, and satisfying results by computer simulation has also obtained.First, the paper analyzed the rotating invariance multiplexed matched filtering recognition method of cascade modified filter theoretically, discussed modified filter with identifying gray target image under background noise or not, and came to the conclusion that the optimization filter is inverse filter when without background noise. Second with inverse filter, the correlating peak of identifying target could be acquired after 30% input of identifying gray target image, while the rotating invariance recognition between 0?and 32?has also been realized. As the spring of correlating peak value between 0?and 4?rotating is concerned, the paper indicated the reason is the loss of gray image value. Third, two cases have been considered to compute the correlating peak value and noise value, that is, when Gaussian noise has been added to the whole background and to the target. At last, the paper discussed the peak value when target is very close to non-target using the neural network method.