Image analysis for plant factory-triangulation of two-dimensional object for polygonal approximation

Abstract In this paper, a triangulation method using edge lines and central lines of two-dimensional object for polygonal approximation is described. A polygonal approximation is one of method of modelling for range images. This technique was applied to the machine vision system of a harvesting robot which can recognize an object spatially. In cases of complex objects, it was difficult to calculate the approximation process. This problem is caused by existence of the unrelated edge line between the edge line and the highest point, and as long as plural edge lines connect to only one highest point, this problem is never resolved. As a solution, we used plural target points given by thinning process instead of only one highest point. Thinning process give a central curve data of two-dimensional object. However, this center curve data is not suitable for target points because it is a large number of sequential point data. Therefore, we performed linear approximation toward central curve, and reduced the central curve data to a suitable number. By using this triangulation method, triangle's overlapping could be avoided and rational triangulation could be done.