Development of 3-D Visual Model Construction System Using Rotary Vision System and Motion Picture Analysis Method

许多视觉系统已经开发并应用于实际生产线，并证明了它们的有效性。但大多数病例都是二维视觉系统。在许多情况下，工业应用的3D视觉系统使用复杂的光学系统，需要专家进行精确校准。为了使三维工业视觉系统更具实用性和易用性，需要开发出适用于工业应用的通用、健壮的三维测量方法。为满足这一要求，基于旋转视觉系统的三维视觉测量方法应运而生。主要完成了以下工作：(1)光学系统：该旋转视觉系统在一台CCD摄像机前安装了一个旋转镜，用于观测小物体的多方向图像。移位后的光学系统可以观察到无失真的图像，并进行精确的三维测量。(2)速度：可以获得运动图像。(3)图像处理算法：图像处理步骤如下：a)多阈值二值化提取物体的正确轮廓。B)在运动图像之间，分析对应的轮廓，并将它们作为同一组进行掩蔽。C)在对应的轮廓线之间，选择对应的特征点。C)在旋转图像中，对应点的轨迹描述圆。这些圆的直径给出了对应点的高度，经过运动图像分析，从物体的三维形状中得到对应点的三维数据。

Many vision systems have been developed and applied into real production lines and proved their effective-ness. But most of the cases were the 2-D Vision systems. In many cases, the 3-D vision systems for industrial applications employ sophisticated optical systems and require precise calibrations by experts. It is required to be specially designed for every application.To make 3-D industrial vision systems more practical and easy to use, it is required to develop general purpose and robust 3-D measurement method for industrial applications. To meet with the requirements, 3-D Visual Measurement Method Using Rotary Vision System has been developed. The following developments have been achieved.(1)Optical System : The rotary vision system employs a rotational mirror in front of a CCD camera and observes multi-directional images of small objects. The shifted optical system observes no distortion images and it makes precise 3-D measurement.(2)Speed : The image can be obtained as motion pictures. The speed of taking pictures is 16 pictures/sec.(3)Image Processing Algorithm : The image processing steps are as follows ; a)Multi-threshold binarization extracts proper contours of objects. b)Between the moving pictures, corresponding contours are analyzed and they are masked as same group. c)Between the corresponding contours, corresponding feature points are selected. c)In the rotational images, the traces of the corresponding points describe circles. The diameters of the circles give the height of the corresponding points.After the motion picture analysis, the 3-D data of the corresponding points from the 3-D shape of the objects.