Joint Transform Correlator (JTC) is a more advanced optical pattern recognizer. Compared with computer correlation
matched method, it has many advantages, such as parallel processing, high speed and more sampling points. The
bottleneck technique of JTC is how to recognize the distorted targets in cluttered scene. The Maximum Average
Correlation Height (MACH) algorithm applied in JTC is presented. The MACH algorithm has powerful capability of
recognition for distorted targets (rotation and scale etc.). Through optimizing MACH frequency domain filter and
projecting it to space domain, the MACH reference image can be obtained. This MACH reference image clearly records
various distortions of a target. Based on the basic principle of JTC, the distorted targets in cluttered scene can be detected.
Optical experiments with JTC show that the edge features of detected image should be extracted, the correlation
calculation of MACH reference image and detected target can be performed accurately and the bright correlation peaks
can be obtained. In the process of optical correlation recognition, the performance of MACH reference image is used
sufficiently. The energy of correlation peaks is greatly enhanced, the cluttered scene can be suppressed effectively and
the MACH reference image has high distortion tolerance. To demonstrate the feasibility of MACH reference image, the
scale distorted car in cluttered scene is successfully detected.