A Study Of 3D Reconstructing From Straight-Line Optical Flow Based On Ant Colony Algorithm

It is a research hotpot in computer domain to carry on 3D reconstruction of rigid body's motion and relative depth after extracting surface features of the 3D rigid body from continuous image sequence. This paper mainly studies on the computation theories and methods of optical flow reconstructing the 3D rigid body's motion and construction on the basis of straight line features.This paper first selects 2D straight line and 3D straight line's expression methods and defines the derivative of the parameter of 2D straight line to the time as the optical speed of straight line, defines the derivative of optical speed as the optical acceleration of straight line, which deduces the motion relationship between the rotation motion parameters of 3D rigid body surface straight line and its straight line optical parameter on the projection plane under perspective projection model, namely, the linear optical flow equations. Base on this equation, the angular velocity and acceleration parameters of the 3D rigid boy can be solved by the linear equations as long as tracking the two straight line optical flow of three consecutive images stably, then solving the parameters of translational velocity, relative depth information of rigid body, cameral focal length and so on.The computation methods of reconstructing 3D motion and construction with straight line optical flow based on ant colony algorithm was presented in this paper, which takes the error of 3D rigid body rotation motion parameter as the target function of ant colony algorithm, takes 3D rigid body surface straight line's parameter on the projection plane and the angular velocity and acceleration parameters of the 3D rigid boy as the input and output of ant colony algorithm model. The method adjusts the input parameter ant colony algorithm model and the cycle-index of ant search continuously to minimum the value of the target function, when the output of ant colony algorithm model is the optimal solution of 3D rigid rotation motion parameter, then solves the point-to-point speed parameter of the rigid body and the space linear coordinate and realized the 3D reconstruction of rigid body. As long as the method can gain and track two plane straight line parameters steadily, then the reconstruction of rigid body rotation, translational motion and elative depths information can be achieved. Multiple sets of simulation experiments show that the system is stable with good robust performance and the counting error.This paper also compares with several other common optimization solution method, such as the genetic algorithm, the grain of subgroup algorithm, the neural network, linear neural network and so on, they were separately used in the straight line equation model to carry on optimization and solution in this paper, and makes detailed comparative analysis with ant colony algorithm in the aspects of computational accuracy and efficiency and so on.