Object Identification Based on Deformable Templates and Genetic Algorithms

In this paper, an algorithm is described which detects and localizes shapes in grayscale images. Deformable templates are often employed to model the geometry of the object to be found in an image. There are several methods to find the optimal placement and deformation of the template onto the image. A genetic algorithm is designed which takes the object template and processed grayscale image and locates the object in the image, invariant to rotation, translation and scale. The limitations of processing the image before running the Genetic Algorithm (GA) are discussed.