Real-time Distortion Invariant Optical Pattern Recognition Reliability

Digital-optical hybrid image processing, Optical pattern recognition (OPR) technique performs image processing and recognition through optical correlation and digital processing. It has promising application prospect in automatic target recognition because of its outstanding merits of high-speed parallel operation, high spatial resolution, flexibility and accuracy. The paper is aimed at improving the recognition reliability of the real-time distortion-invariant Optical Pattern Recognition system to meet the requirement of its practical application.By comparing the image-preprocessing results of different processing algorithms, an optimized algorithm was decided to ensure satisfying correlation SNR and preprocessing rate to be obtained. Synthetic Discriminant Function (SDF) filters of the 3-D targets were designed and the Reference Filter Libs (RFL) were constructed to provide high correlation SNR. Real-time distortion-invariant optical correlation was realized by temporal-spatial multiplexing technique. A kind of Artificial Neural Network (ANN) was used to perform the correlation signal postprocessing. As a result, the recognition possibility of the OPR system was improved efficiently compared with the conventional methods.From the objects recognition tests buried in strong noise background, we found even with the ANN post-processing technique, a certain amount of recognition errors (about 10%) were still unavoidable. Therefore, based on the ANN correlation signal post-processing technique, we designed a kind of SDF performed in spatial domain, and a two-step recognition technique was introduced to perform real-time distortion-invariant recognition by combining frequency domain filtering of SDF with spatial domain filtering of SDF. This is the originality of the paper. The experimental results of hardware optical implementation show that the recognition probability and reliability of the OPR system are greatly improved (about 99%). This provides strong technique support for the practical and engineering application of the OPR system.