Abstract

Based on the results obtained, it can be concluded that the choice of the parameters (M), (T) and (S_f) significantly affects the quality of the detected contours, and that their optimal combination depends on the algorithm used and the characteristics of the image.

The contrast between the target layer and the base is one of the most determining factors for obtaining contours that accurately represent the specimen geometry. For additively manufactured parts with more complex geometries, the use of cropping masks or techniques such as the Iterative Closest Point (ICP) algorithm may improve contour identification.

Among the algorithms studied, Canny has shown a wide operating range and consistent results, whereas ZeroCross requires additional processing to achieve comparable results. Similarly, image noise must be addressed according to its characteristics, avoiding excessive filtering that could result in the loss of relevant contour information.

On the other hand, automatic methods reduce user intervention, although they are more sensitive to parameter variations and may provide lower accuracy under certain conditions. For methods requiring manual point selection, this intervention introduces a dependence on user accuracy, which can be reduced by using longer cropping masks, although at the cost of increased processing time.

Overall, there is no single optimal algorithm for all situations. Therefore, the selection of the most appropriate method should primarily consider the image contrast, noise, geometric complexity and required level of automation.