Single-View´s 3D Explanation Based on Prior Knowledge and Constraint Satisfaction

In this paper, we estimate the geometric characteristics, such as the relation of edges and the angle characteristics between edges, from single view. And we also use the prior knowledge about objects, such as the feature of edges and the geometry construction of objects, to construct the variable constraint based on the projection rules. In this paper, we transfer the problem of interpretation of 3D objects from single view into a non-linear optimization problem. After local constraints merge into overall constraint, the posture of object can be estimated quickly. Then we can calculate geometry characteristics of the objects for invariant recognition.