Hierarchical search and sequential scene matching in a non real-time envivronment

Includes bibliographical references (pages 61-62)A study of classification methods including the Bayesian is\ud presented. Sequential classification is basically non-Bayesian,\ud and is therefore specially suited for problems in pattern\ud recognition, which requires complex computations.\ud An algorithm was developed to implement hierarchical search.\ud The implementation was done in Double Precision. Since the Rockwell\ud 6502 Development System in which the implementation was done is a\ud single precision, the double precision capability had to be written\ud into the program.\ud The processes involved in the hierarchical search require\ud operations in digital filtering and convolution.\ud The algorithm proposed was written for a non real-time\ud application because several sequential processes were used. The \ud first process involved geometric correction. If the window has a\ud different orientation from the object located in the search region,\ud several re-orientations of the window may be required before the\ud identification of the object can take place