Proposal of visual servoing using phase-only-correlation (POC)

Image-based visual servo using coordinates of feature points has been well discussed and used in many applications. However, these methods rely on the perfect matching of feature points, which is often difficult in an actual situation. Recently, visual servoing methods based on region informations such as mutual information and an image transformation matrix are proposed. While these methods can achieve more robust visual servoing, there is still some difficulties in how to extract and apply such information. In this paper, region based approach using Phase-Only-Correlation (POC), which is known as a subpixel image matching algorithm, is proposed. Although POC can be applied only in 4 DOF (degree of freedom) without X and Y axis rotation, it accomplishes robust and accurate positioning. We also reveal that the image jacobian, which is often difficult to estimate in realtime, can be written as time invariant matrix in 4 DOF with proposed control scheme.