Three-dimensional shape construction and recognition by fusing intensity and structured lighting

摘要

In this paper, a new technique is developed for three-dimensional (3D) shape reconstruction and recognition from two-dimensional (2D) images of multiple objects with mutual occlusion. To facilitate such an analysis, both intensity and structured-light coded images are used. From intensity images, equi-brightness contours, which are swaths of points sharing similar intensities, are located. It is shown that equi-brightness contours possess distinct shape and orientation for different types of surfaces, and hence, can be used to distinguish them. To achieve a quantitative description of the shape and orientation of the imaged surfaces, the projected pattern in the corresponding structured-light coded images is analyzed. A scheme is developed which relates the orientation and curvature of the projected pattern observed in the structured-light images to the surface orientation and principal surface curvatures of the imaged objects.