Vision System for Environment-Driven Robots using Chipped Range Image Sensor

我们提出了一种利用距离图像实时跟踪物体三维运动的新方法。非常高速的测距传感器“硅测距仪”可以获得超过每秒30帧的测距图像，并旋转以保持物体在可测量区域。粗略而快速的距离图像处理结果控制着距离传感器的角度，以目标为中心的距离图像决定目标的姿态。图形工作站的三维CG渲染硬件可以快速地从表面模型生成距离图像，从而简化了比较算法。对于人机协作系统，估计运动的跟踪速度和精度似乎是可行的。测量自然环境中自由形状物体的运动是一个非常困难的问题。对于刚体物体的实时姿态估计问题已经有很多研究者进行了研究，但是能够实现全三维姿态估计的系统很少。为解决这一问题，研制了先进的无接触测量测距传感器。虽然通常的测距仪是非常昂贵和缓慢的，但现在我们知道一些非常高速的测距仪使用大规模集成电路技术。硅测距仪是本实验室研制的高速测距传感器之一。该测距仪可测量24^<**>24像素范围内的图像，每秒30帧以上，精度在1毫米以下。在此基础上，提出了一种新的距离图像实时处理方法。“综合分析”的概念以复杂和不可分析的问题而闻名。此外，这种方法非常灵活，因为“合成”过程是所有应用程序的共同过程。为了与实测距离图像进行比较，必须生成多个距离图像，因此需要快速生成算法。因此，我们提出了一种利用图形渲染硬件生成距离图像的新方法。图形硬件的开发是为了快速生成强度图像，但使用Z-Buffer也可以快速生成范围图像。普通图形工作站有一个带z缓冲区的图形加速器。为了证明我们的方法是有效的，我们选择了一个自由形状物体的6自由度跟踪任务。快速的距离图像合成使得对曲面物体的实时跟踪成为可能。该方法也可应用于其他距离图像的处理。少

We propose a new method to track 3-D motion of an object in real-time, using range images. Very high speed range sensor "Silicon Range Finder" makes it possible to acquire range images more than 30 frames per second, and rotated to keep the object in measurable area. Rough but fast result of range image processing controls the angle of the range sensor, and the object centered range images decide the pose of object. 3-D CG rendering hardware of the graphics workstation generates range images from the surface model very fast, so it makes comparison algorithm simple. Tracking speed and accuracies of estimated motion seem to be feasible for the man-machine cooperative systems.It is a very difficult problem to measure a movement of free-form object in natural environment. The problem of real-time pose estimation of a rigid object has been approached by many reseachers, but there are few systems capable of full 3-D pose estimation.To solve this problem, range sensor is developed for advance … More d measurement without contact. Though usual rangefinder was very expensive and slow, but now we know some very high speed rangefinders using LSI technology. Silicon Range Finder is one of high speed range sensor developed in our laboratory. This rangefinder can measures 24^<**>24 pixels range images more than 30 frames per second, and its precision is under one millimeter. Then we propose a new real-time processing method for the range image.The "Analysis by Synthesis" concept is known for complex and not analyzable problem. Also this method is very flexible, because the "Synthesis" procedure is a common process to all application. Several range images must be generated to compare with measured range image, so the fast generating algorithm is needed. Therefore, we propose a new method to generate range images using graphics rendering hardware. Graphics hardwares are developed to generate intensity images fast, but range images can be also generated fast using Z-Buffer. Ordinary graphics workstation has a graphics accelerator with Z-buffer.To prove our efficient method, we selected a task that is 6-DOFs tracking of free-formed object. Fast range image synthesis makes it possible to track a object with curved surface in real-time. Also our method can be applied to other range image processing. Less