A Visual Servoing Strategy on Multiple Triangles using the Image Moment Approach

Abstract

In this work we propose a visual servoing strategy on multiple triangles underoing different piecewise rigid motion. Our strategy is to compute first a correction related to each triangle independently using its related interaction matrix. Then, we process an average global correction to correct the camera pose and obtain a view which is as close as possible to the desired visual features. We show that this strategy performs better than a naive one with a single interaction matrix that gathers all the triangles. We use an analytic version of the image moments on 2D triangles and apply the proposed strategy on several triangle poses to show its relevance.