Marker-Based Shape Estimation of a Continuum Manipulator Using Binocular Vision and Its Error Compensation

Robot-assisted laparoscopy endoscopic single-site surgery robot (LESS) procedures typically involve highly flexible and continuum manipulators to get access to the surgical targets through a small keyhole incision and then perform complicated operations within the abdomen or chest cavity. However, the inherent deformable design and inevitable collisions with the anatomy cause both active and passive deformation of the distal continuum manipulator, which challenges the accurate shape estimation in both structured and unstructured environment. Therefore, it is essential to achieve real-time and accurate and shape tracking of the continuum manipulator for precise and reliable motion control. A new method is proposed in this paper for solving in the problem of real-time shape sensing for the distal continuous robot using binoculars by extracting position information of each markers. However, to cope with the problems of visual occlusion and limited algorithm stability, an error compensation method based on machine learning algorithm is realized. The performance of the proposed algorithm was verified by experiments, and the accuracy of the reconstruction after the correction by the learning algorithm was improved by about 70%.