Abstract: In this paper we present a method of segmentation usinga multiagent system, and an application to fish otolithgrowth ring detection. The otoliths images are composedof alternative concentric dark and light rings, thenumber of which increases with the age of the fish. Upto now, the identification of growth rings, for ageestimation, is routinely achieved by human readers, butthis task is tedious and depends on the reader'ssubjectivity. The system proposed here is composed ofseveral agents whose individual task is to detect localextremes on a grayscale image. For this aim the agentsare provided with sensors on the gray levels of theimage. By computing the mean gray level of two sensorsplaced in front of it, the agent, if it searches forlight rings (respectively dark) will decide to turn inthe direction of the lighter (respectively darker)sensor. The path of the agents has been tested as aroof edge detector, using the Canny criteria: gooddetection, good localization, and low multipleresponse, in order to choose the best parameters rulingthe agents behavior, according to the image structures.Tests have been first achieved on synthetic images, andthen on otoliths images. !11
展开▼
