This work is concerned with one of the methodologies used in the final stages of machine vision: the matching of model point patterns to observed point patterns. Conventional search methods not only fail to arrive at the optimal match, but are also computationally expensive and time consuming. To arrive at the optimal pattern match, stochastic and heuristic optimization as the search technique, exploiting Simulated Annealing (SA), Evolutionary Programming (EP) and Mean Field Annealing (MFA), are explored in detail. A comparison of results obtained using SA versus "hill-climbing" and "exhaustive search" techniques, and results of EP are presented. The relative effectiveness of these optimizing search algorithms over other conventional algorithms will be demonstrated. Finally, the limitations of MFA are discussed.
If you have any questions please contact the ETD Team, libetd@njit.edu.