Autofocus for enhanced measurement accuracy of a machine vision system for robotic drilling

Abstract

Erroneous object distance often causes significant errors in vision-based measurement. In this paper, we propose to apply autofocus to control object distance in order to enhance the measurement accuracy of a machine vision system for robotic drilling. First, the influence of the variation of object distance on the measurement accuracy of the vision system is theoretically analyzed. Then, a Two Dimensional Entropy Sharpness (TDES) function is proposed for autofocus after a brief introduction to various traditional sharpness functions. Performance indices of sharpness functions including reproducibility and computation efficiency are also presented. A coarse-to-fine autofocus algorithm is developed to shorten the time cost of autofocus without sacrificing its reproducibility. Finally, six major sharpness functions (including the TDES) are compared with experiments, which indicate that the proposed TDES function surpasses other sharpness functions in terms of reproducibility and computational efficiency. Experiments performed on the machine vision system for robotic drilling verify that object distance control is accurate and efficient using the proposed TDES function and coarse-to-fine autofocus algorithm. With the object distance control, the measurement accuracy related to object distance is improved by about 87 %.