Visual navigation technology for intelligent vehicle - Road model for autonomous driving and alarming

一些研究人员利用各种视觉信息源，包括颜色、视差、距离、光流等，研究了自动道路车辆的视觉引导导航。这些研究强调了确定前方道路作为智能道路车辆发展的重要组成部分的重要性。在这个项目中，我们提出了三个模型来解决智能车辆的道路跟随和3d道路形状重建的一般问题。以下三种方法基于活动轮廓模型（ACM）。1)该方法假设道路边界平行，道路宽度恒定。然后，我们使用受并行性约束的耦合活动轮廓模型检测和跟踪图像中的道路区域。然后，该系统从一张图像中生成一个3d道路模型。我们通过应用于包含4000多张图像的真实道路场景来评估该方法的有效性。2)该方法是基于全向感知的反应式视觉导航。全向图像传感器HyperOmni Vision将机器人投影在输入图像的中心。因此，可以通过活动轮廓模型提取机器人周围的粗糙自由空间。该方法生成低级命令，使机器人保持在自由空间的中间，并通过平衡提取的活动轮廓的形状来避免碰撞。机器人通过主动轮廓模型跟踪闭环曲线，避开障碍物，沿走廊移动。此外，该方法可以在图像坐标上直接表示环境与机器人之间的空间关系。因此，该方法可以在不进行几何三维重建的情况下对机器人进行控制。3)第三个模型用于机器人沿路线导航。当机器人移动到目标位置时，地平线上一系列连续的全方位图像会记住路线。在机器人导航到目标点的过程中，利用双活动轮廓模型将输入与记忆的时空路径模式图像进行匹配，从而估计出机器人的准确位置。少

Several researchers have investigated visually-guided navigation of autonomous road vehicles using various sorts of visual information sources, including color, disparity, range, optical flow, etc. Such research has emphasized the importance of determining the road ahead as a significant component of the development of intelligent road vehicles. In this project, we propose three models for solving the general problem of a road following and 3D-road shape reconstruction for a smart vehicle. Three following methods are based on active contour models (ACM).1) The method assumes that road boundaries are parallel and that the width of the road is constant. We then detect and track the road region in the image using coupled active contour models subject to a parallelism constraint. The system then generates a 3D-road model from a single image. We evaluate the effectiveness of the method by applying to real road scenes comprising more than 4000 images.2) The method is for reactive visual navi … More gation based on omnidirectional sensing. The robot is projected at the center of the input image by the omnidirectional image sensor HyperOmni Vision. Therefore, rough free space around the robot can be extracted by active contour model. The method produces low-level commands that keep the robot in the middle of the free space and avoid collision by balancing shape of extracted active contours. The robot can avoid obstacles and move along the corridor by tracking the close-looped curve with an active contour model. Furthermore, the method can directly represent the spatial relations between the environment and the robot on the image coordinate. Thus the method can control the robot without geometrical 3D reconstruction.3) Third model is for navigating the robot along a route. The route is memorized by a series of consecutive omnidirectional images at the horizon while the robot moves to the goal position. While the robot is navigating to the goal point, the input is matched with memorized spatio-temporal route pattern images by using dual active contour models and we can estimate the exact robot position. Less