Robust Line-Segment Extraction using Genetic Algorithms

Success in scene interpretation in high level computer vision depends heavily on the quality of features derived from the low level stages of processing. We describe an optimisation process for robust low level feature extraction based on Genetic Optimisation. The fitness function is a performance--Momeasure which reflects the quality of an extracted set of features. We shall present some results and compare them with a Hill-Climbing optimisation approach.