A Simple Global Calibration Method Based on 1D Target for Multi-binocular Vision Sensor

Abstract

To solve the problem of global calibration for multi-binocular vision sensor in a wide measuring range, a method based on 1D target is presented. The 1D target is only required to be put in a number of proper positions between two vision sensors so that one sensor can observe one end feature point and the other sensor can observe the other end point. These two sensors need not to have common vision field, and motion of target need not to be known. The rotation matrix and translation vector between sensors are estimated by nonlinear algorithm optimizing the cost function based on the distance constraint between two extremity feature points of 1D target. Both of simulation result and real data prove that the global calibration method proposed is valid, simple and practical, and has relatively high calibration accuracy.