By extracting the basic rectangular features of an object from a digitized Fourier transform image, one can calculate the scale factor and rotational angle of a rectangular object and use them to derotate and scale the input image. Thus, a scale- and rotational-invariant optical correlator is achieved. The accuracy of the computed scale and orientation parameters yielded by this approach ensures nominal correlator performance for object recognition and tracking applications. The concept and system implementation of such an intelligent optical correlator are discussed, and the results of performance analysis are presented.