Research On The Iterative Optimization Method In Computer Vision

Vision measurement system has a large number of optimization problems. Solving these optimization problems has important significance to enhance and improve the measurement accuracy. This paper mainly researches optimization of the camera calibration and3D coordinate calculation.This paper studies the common optimization methods in computer vision, and analyzes both advantages and disadvantages of the method. Based on the existing camera model, we use nonlinear camera model with distortion parameters to build an optimization model for camera calibration, and design an iterative optimization method. This method is based on the two steps theory. Firstly, without considering the distortion, the intrinsic and extrinsic parameters of initial value is estimated linearly. Then, the initial value of distortion coefficient is solved by step through the nonlinear optimization method, and both the internal and external parameter is optimized. Finally, all the camera parameters are obtainedThis paper uses nonlinear method to solve the3D coordinates of the feature points. According to problem that the ray with corresponding image point could not intersect at one point, the optimization model is established with the minimum re-projection error, and the iterative optimization method also be designed for this model. After the linear initial estimates of the3D coordinates is obtained based on camera calibration, the camera parameters and the3D coordinate is optimized iteratively according to the thought of bundle adjustmentThrough the vane experiment, crafts sculpture experiment, and automobile parts experiment, the feasibility of the optimization method is verified in solving the above optimization problem, and the accuracy of measurement is effectively improved.