Aesthetics Evaluation for Robotic Chinese Calligraphy

摘要

A great many efforts have been devoted to reproducing a reference image for Chinese calligraphy with a robot. However, there exist modeling errors in mapping from the presentation of a reference image to brush trajectories. In this paper, a closed-loop calligraphy system is established to decrease effects brought by modeling errors and optimize aesthetics. A strategy of robotic Chinese calligraphy is newly proposed based on aesthetics evaluation, which is characterized by three aesthetics indexes. Each stroke is planned online to achieve the best aesthetics effect of the whole calligraphy work. A constraint optimization problem is formulated to optimize the aesthetics effect, resulting in the trajectory of each stroke for robotic control. Moreover, the concept of the complete feature set is presented to describe ideal features, which simultaneously characterize a calligraphy task as well as calligraphy aesthetics. The rationality of the complete feature set is analyzed in both qualitative and quantitative way. The advantage of the proposed strategy is verified in a simulation of robotic Chinese calligraphy on a database containing 400 calligraphy references. To validate effectiveness of the proposed theory, Chinese calligraphy is performed on an NAO robot with and without the proposed aesthetics evaluation.