Autonomous Viewpoint Selection of Robots Based on Aesthetic Composition Evaluation of a Photo

摘要

In the field of painting and photography, composition rules are widely used under various circumstances in order to arouse humans' aesthetic evaluation. In this paper, we propose a method to evaluate the aesthetic values of a scene according to some certain composition rules, such as Rule of Third, Diagonal Composition and Triangle Composition. We propose an evaluation function with three factors for each kind of compositions, which is reasonable to describe properties of compositions as the result showed in our questionnaire. We develop an observation position searching method by estimating relationships between targets from different viewpoints of the monitoring robot. Then a score will be obtained for each viewpoint using our evaluation function. By these scores, the best observation position can be determined within a reachable field. With path planning and moving control, the monitoring robot arrives at the best observation position. We suppose our research can provide any hint about common understanding between human beings and robots.