An Approach to Perception Enhancement in Robotized Surgery Using Computer Vision
Agustin A., Albert Hernansanz, Joan Aranda, Alı́cia Casals · InTech eBooks · 2010
The capacity of the human visual system to perceive 3D space from a monocular view is characterized by a set of stimulus and processes. As human's perception of depth on a flat surface is enhanced by the use of linear perspective, a set of geometric properties from the projective transformation provide enough 3D information to effectively calculate the pose of an object from 2D images. In this work, a model of the variance of a geometric configuration under perspective transformation has been developed. Particularly, this geometric configuration is the angular variation caused by perspective distortion. Since the resulting projected variation depends on the position of the observer, the exterior orientation of the camera can be determined by the model. In the presented approach the image plane contains the whole required information to solve the exterior orientation problem. Thus, there is no necessity to measure angular samples from determined 3D structures. Experimental results show the efficiency and robustness of the method, in this case using a set of concurrent lines. Its accuracy can be affected by the reliability of the line feature