Development of Driver Assistance System Based on Advanced Usage of Multiple View Images

通过本研究，获得了以下结果：1。提出了HIR （Human-Oriented Information Restructuring）系统。HIR系统对多视图图像等各种信息进行整合和重构，并将重构结果以图像形式显示给用户。作为HIR的一个例子，“鸟瞰系统”已经开发出来，它可以从十字路口上方生成任意视点图像。自由视场生成算法基于二维投影变换和汽车高度估计。该系统以1/24的立体模拟方式生成大小为160 × 120，帧/秒的鸟瞰图。研制了一种十字路口右转情况下的视觉辅助系统。该系统利用安装在对面汽车和基础设施上的摄像头图像，产生视觉辅助图像，为驾驶员消除死角。一辆能在真实环境中工作的原型车已经研制出来了。“我的车”配备了5台电脑、3个摄像头、一个GPS传感器和三轴方位传感器。“对立车”配备了电脑、摄像头、GPS传感器和三轴方位传感器。交通信号灯上安装了基础设施摄像头。开发了一种能够在右转情况下生成驾驶员辅助图像的视觉辅助系统，并在真实环境中进行了评估。结果表明，驾驶员对对面车辆的识别是正确的。由此得出结论，多视点图像的先进利用对驾驶员辅助系统是非常有益的。

Through this research, the following results have been obtained :1. The HIR (Human-Oriented Information Restructuring) system has been proposed. The HIR system integrates and restructures various information such as multi-view images, and displays the results of restructuring to the users in an image form.2. As an example of HIR, "Bird's-Eye View System" has been developed, which generates an arbitrary viewpoint image from above an intersection. The algorithm of the free-view generation is based on 2-D projective transformation and estimation of the heights of a car. This system generates bird's-eye view images in the size of 160x120, 10 frame/sec, in a 1/24 diorama simulation.3. A visual assistance system in a right turn situation in an intersection has been developed. This system utilizes camera images installed on the opposed car and infrastructure and generates visual assistance images which eliminate blind corners for the driver.4. A prototype car that works in a real environment has been developed. "My Car" is equipped with 5 PCs, 3 cameras, a GPS sensor, and tri-axis orientation sensor. "Opposed Car" is equipped with a PC, a camera, a GPS sensor, and tri-axis orientation sensor. An infrastructure camera is installed on the traffic signal.5. A visual assistance system that generates driver assistance images in a right turn situation is developed and evaluated in a real environment. The results show that the driver recognizes the opposed car correctly.6. From these results, it is concluded that the advanced utilization of multi-viewpoint images is very useful for driver assistance system.