Abstract: The joint transform correlator and Vander Lugt correlator applied in automatic target recognition are introduced in order to allow the optical correlation system to achieve a better application and guarantee the excellent recognizing effectiveness in automatic target recognition field under high-speed condition. Aiming at the key technologies in the realization process of the matched-filter correlator, the establishment of equal correlation peak synthetic discriminant functions to solve the problem appearing in the aberrance invariability recognition is brought forward. Furthermore, BP artificial neural net to recognize the correlation peak in correlation plane is put forward to solve the deficiency of threshold method in the correlation peak recognition technique. The nerve cell number of input layer is reduced and the ANN structure is simplified by dividing the interesting regions. The optical correlation system built in the lab proves the method can classify the correlation peaks from noises effectively and can improve the reliability remarkably.