基于单模板的二维场景重建方法

针对现有场景重建算法具有计算复杂度高且鲁棒性差的缺点,本文采用一种基于映射变换的黑箱标定方法,根据图像坐标与世界坐标的线性映射模型,利用矩阵方程获取摄像机参数。与传统方法相比,该方法仅需一幅模板图像,降低了特征点与模板个数,简化了计算。实验过程中,利用图像降噪、改进的Harris算子角点提取等预处理过程快速提取特征点,实现摄像机参数标定的映射变换;建立了误差评价函数评价本文方法与Tsai标定算法。实验结果表明,该方法具有稳定性强、计算精度高、算法简捷的特点。

2D view reconstruction method based on single calibration pattern

In order to solve the high complexity and the insufficient robustness of existing reconstruction methods, a black-box camera calibration method on the basis of the mapping transformation was presented. According to the linear mapping model between the image-plane reference frame and the object-space reference frame, the matrix equation to calibrate the camera was used. Comparing with the traditional methods, the method only needs one plane-model image to reduce the quantity of the feature points and the template, and simplifies the calculation. In the experiment the pre-processings, such as image filtering and Harris operator corner detector, etc. , were used to get feature points quickly and realize the mapping transformation of camera parameters. The error estimate function was established to compare the proposed method with Tsai calibration algorithm. The experiment results show that the method is characterized by good stability, high precision and algorithm simplicity.