The goals of this project are to formulate, design and implement evolutionary algorithms for a wide range of problems in computer vision. Evolutionary computation encompasses a variety of population based problem solving techniques that mimic the biological process of Darwinian evolution which is based on the principle of natural selection. Evolutionary algorithms provide powerful, yet versatile problem solving mechanisms for search, adaptation, learning and optimization in a variety of application domains. This project investigates the genetic algorithm (GA), an important member of the wider class of evolutionary algorithms, in the context of computer vision problems such as edge detection, image segmentation and feature extraction.
 
    The problems of edge detection, image segmentation and feature extraction are modeled as combinatorial optimization problems that entail the minimization of an appropriately defined cost function. Chromosomal representations, fitness functions and GA operators are formulated for each of the above problems. The shortcomings of the traditional GA approaches in the context of the above problems are exposed. This project proposes a class of hybrid evolutionary optimization algorithms that combine the strengths of annealing based techniques, such as simulated annealing, micro canonical annealing and random cost search with those of the GA while alleviating their individual weaknesses resulting thereby in performance that is superior to that of either class of techniques used in isolation.
 
 
Publications
 
S.M. Bhandarkar, J. Zeppen and W.D. Potter, A Genetic Algorithm for Line Feature Extraction, Proc. Intl. Conf.  Ind.  and Engr. Appl. AI and Expert Sys., Castellon, Spain, June 1-4, 1998, Vol. 1, pp. 647-656.
 
S.M. Bhandarkar and X. Zeng,  Figure-Ground Separation: A Case Study in Energy Minimization via Evolutionary Computing, Proc. IAPR Intl. Wkshp. Energy Minimization Methods in Computer Vision, Venice, Italy, May 21-23, 1997, pp. 375-390.
 
S.M. Bhandarkar and X. Zeng, Evolutionary Computation for Figure-Ground Separation,  Proc. Intl. Conf. Neural Networks, June 9-12, 1997, Houston, TX, Vol. 3, pp.  1673-1678.
 
S.M. Bhandarkar and X. Zeng, Evolutionary Approaches to Figure-Ground Separation, Applied Intelligence, under review.
 
S.M. Bhandarkar, Y. Zhang and W.D. Potter, An Edge Detection Technique Using Genetic Algorithm-based Optimization, Pattern Recognition, Vol. 27, No.  9, Sept. 1994, pp. 1159-1180.
 
S.M. Bhandarkar and Hui Zhang, Image Segmentation Using Evolutionary Computation, IEEE Trans. Evolutionary Computation, accepted for publication.
 
S.M. Bhandarkar, Y. Zhang and W.D. Potter, A Genetic Algorithm-based Edge Detection Technique,  Proc. Intl.  Joint Conf. on Neural Networks, Nagoya, Japan, October 1993, pp. 2995--2999.

S.M. Bhandarkar, Y. Zhang and W.D. Potter, Detecting Edges with a Genetic Algorithm, Proc. Intl. Conf. Artificial Neural Networks in Engineering, St. Louis, MO, November 1993, pp. 357-362.