Range estimation and object identification with a single camera machine vision system

Typically machine vision systems estimate ranges using stereoscopic vision systems or specialized range detectors. Although these methods are effective, they are expensive. In this paper, a single camera range estimation algorithm is developed for a low-cost machine vision system that uses off-the-shelf components. This algorithm uses the perspective transformation and a priori knowledge of scene’s objects. The perspective or imaging transformation maps a point, ( X, Y, Z), from the three-dimensional world coordinate system into an image point, ( u, v, w), in a two-dimensional image plane. Assuming that the single camera’s optical axis lies along the w and Z axes as shown in Figure 1, the image plane lies at ( u, v, 0). In the Cartesian coordinate system, the perspective transformation is nonlinear. However, the perspective transformation can be linearized by mapping Cartesian coordinates into homogeneous coordinates. A point, w, with Cartesian coordinates, ( X, Y, Z), has the homogeneous coordinates