Segmentação automática de contornos em imagens de peixes e modelagem tridimensional para caracterização morfológica

Abstract This work proposes a methodology, using digital image processing concepts, for the automatic segmentation of contours in fish images, thus contributing to subsequent studies focused on detecting variations in their body shape. These morphological or morphometric studies are generally performed for the management of fishery production, and its importance is that artisanal fishing is one of the main economic activities in Brazilian communities. For this purpose, thresholding and border detection techniques (in particular, the Canny algorithm) were used, which will be validated with a similarity criterion based on the Hausdorff distance calculation. In addition to segmentation in two-dimensional images, a method for representing organisms in three-dimensional form is also proposed, using the Structure Sensor technology for 3D scanning. The results showed that the methodology obtained good results, with a mean similarity of 98.38 %, but the method was not effective in images that were not acquired correctly.