Fault Tolerance Reconstruction for Multiple Moving Objects Using Reduced Camera Set

Using image information to make 3D model reconstruction of real objects in computer vision and computer graphics is getting more and more attentions. Widely used methods include reconstructing object based on the multiple silhouettes and stereo vision. In this work, we design a fault tolerance system for multiple moving object reconstruction using three cameras located in the Cartesian coordinate frame. In the first stage, digital image processing (DIP) is used to extract vertexes and edges. In the second stage, we apply the principle of perspective projection to analyze the coordinates of features in the 3D space, and then iteratively remove the redundant vertices by reasoning. We also develop an adaptive method to prevent any avoidable misjudgement and sudden transition.