A high-precision binocular method for model-based pose estimation

It is an important problem to determine the relative position and translation between two space objects in robotic research. However, if traditional stereo vision method is used in model-based system without exploiting rigid constraints, the precision may be low especially in depth direction. This paper solves this problem with an accurate and real-time strategy, which is based on real-time genetic algorithm (RGA) and perspective 3-point (P3P) algorithm. Real image and simulation experiments show this method has dramatic performance and it can be applied in real-time situation.