Accurate dynamic vision system calibration using circular control points

Abstract

Dynamic vision system (DVS) is a high time resolution camera. Due to the lack of standard frame pipeline, DVS cannot directly detect traditional control points (such as chessboard corners and circle centers). Two methods based on blinking pattern and calibration panel motion are proposed in this paper to solve the problem. The detection accuracy of blinking circular control points on the Light-Emitting Diodes (LED) is greatly improved by the proposed event frame convertor and the root mean square (RMS) is less than 0.04. Besides, this paper presents an incomplete grid detection method to increase the detection success ratio (0.97) and an effective event frames selection strategy to decrease the RMS (0.15). The results show that the standard deviation of both methods is substantially reduced compared to the conventional methods.