基于直线光流场的三维检测新方法

New method of 3D detection based on straight-line optical flow

提出了基于直线光流场的三维检测计算方法。在透视投影下,将表示二维直线的参数对时间的导数定义成直线光流场,详细推导出基于直线光流场检测三维刚体运动参数的模型。使用最小二乘法,求出刚体的三维旋转运动参数、平移运动参数和空间直线坐标,从而实现了刚体的三维检测。该方法利用单目图像序列中连续3幅图像的2对直线光流,能够检测出物体的运动参数和结构参数,有助于全面检测和识别物体。多组实验表明该方法是稳定的,具有较好的鲁棒性能。

In this paper,a new method of solving 3D rigid body motion and structure based on straight-light optical flow is proposed.Under the perspective projection,the derivatives of the three parameters of a 2D line to time are defined as the straight-line optical flow;and the motion parameter model of a 3D rigid body based on straight-line optical flow is established.The rotation motion parameters,translational motion parameters and 3D coordinates of the 3D rigid body can be solved.In order to detect and recognize the details,the presented algorithm uses two pairs of straight-line optical flows of three consecutive image frames in the image sequence and can reconstruct the 3D motion and structure parameters from the straight-line optical flows.Experiments on synthesized image sequences and real image sequences show that this method is stable and robust.