Research of the High-Speed Fusion of Disparity Images from Two Orthogonal Stereo Pairs

本文的研究结果总结如下：(1)将放置在相机旁边的鼻子反射的图像与物体图像同时拍摄，可以获得不受光源照度和颜色影响的图像。参考鼻的表面设计能够反射物体的模糊图像，利用主像旁边的小图像部分对主像中对应的所有像素进行校正。映射受图像深度的影响较小，但可以通过将参考鼻子放置在更靠近镜头位置的位置来减小这种影响。(2)利用立体相机方法实现了平面上各点的深度测量，并通过在曲面上绘制垂直的彩色条纹，实现了对曲面的深度测量。(3)近似平行于视频扫描线的彩色边界线，采用沿彩色边界线同时检测特征点的方法很难检测出来。本文研制了一种视频速率测距仪，用于测量彩色边界线在包括平行边界线在内的任意角度的深度。该探测仪采用一对立体摄像机，连接一对摄像机的每根立体轴彼此成直角，一个水平放置，另一个垂直放置在地板上。所有的信号处理都是通过电子电路进行的，因此无需加载安装在自动行驶车辆中的控制计算机，就可以实现到彩色标记距离的视频速率测量。(4)结合双相机采集的深度数据，提高了深度测量的正确率和采集率等性能。研究奠定了测量的基础和设备的配置，表明该方法在深度测量中是有效的。

The results obtained in this research are summarized as follows :(1) By taking the photography of the image reflected from a nose placed beside a camera simultaneously with an object image, the image can be obtained that is not affected by the illuminance and color of a light source. The surface of the reference nose is designed such that it can reflect the blurred image of the object, and the small image part beside the main image is utilized to the correction of all pixels corresponding the ones in the main image. The mapping is influenced slightly by the image depth but the influence can be reduced by placing the reference nose nearer to the lens position.(2) This study allowed to obtain the depth to every point of the planar surface using stereo camera method, and showed that measurement of the depth also to the curvilinear surface became feasible by painting the vertical color stripes on the surface.(3) The detection of the color boundary lines nearly parallel to video scanning lines is difficult by using the method of detecting the characteristic points along also those lines. A video-rate range finder has been developed to measure the depth to the color boundary lines at any angles including parallel boundary lines. This finder employs a pair of stereo cameras and each stereo axis connecting a pair of cameras is then set at right angle to each other, the one is placed horizontally and the other vertically against the floor. All the signal processing is carried out by electronic circuits, thus enabling the video-rate measurement of the distances to color marks without loading a control computer installed in an autonomous running vehicle.(4) Performances such as the correct and the acquisition ratios in depth measurements have been increased by combining the depth data obtained by the pair cameras. This study established the ground of the measurement and the configuration of the devices, and showed that this method is effective in the depth measurement.