【作者】 孙海涛；

【导师】 陈无畏；

【作者基本信息】 合肥工业大学 ， 车辆工程， 2005， 硕士

【摘要】 视觉导航在智能车辆的智能导航中占有重要地位。由于计算机视觉相对于常用的传感器如测距仪等,具有信息量丰富,智能化水平高等特点,因此其处理的数据较大。同时在复杂或者较危险环境中如何检测路径图像是主要难题。本文选用了标识线图像识别的导航方法。该导航方法原理简单,实现容易,且技术成本和费用低。该方法的主要问题是标识线的变形矫正和边缘检测。本文以两轮驱动智能车辆为研究对象,提出了一种简单可行的图像矫正算法——标定法,能够有效地降低标识线图像变形对参数提取的干扰,为智能车辆导航路径的识别提供了准确的图像信息。在视觉图像处理方面,提出了一种先进行阈值分割,然后再对二值化的图像进行滤波的图像预处理方法,这其中包括改进的最优阈值法、计算量较小的形态学滤波处理方法以及基于二阶导数信息的边缘检测算法。同时在车体相对位置参数提取上进行了改进。仿真中采用了模糊最优控制方法,且仿真计算证明了该控制方法对智能车辆的跟踪控制有着良好的效果。为验证以上图像处理算法和跟踪控制策略的正确性以及智能车辆实际工作效果,在自行研制的智能车辆实验平台上,完成了上位机的主程序(包括动态图像处理、控制算法和数据通讯)以及下位单片机电机控制程序,并且进行了路径跟踪试验。试验结果表明基于以上算法的程序能够满足图像处理的实时性,且智能车辆具有较好的抗噪声干扰能力和较为准确和可靠地对路径进行跟踪。更多还原

【Abstract】 The application of vision-based systems to an intelligent vehicle is a very challenging task. The challenge arises from both the methodologies to handle the multiple problems caused by the structure of the vision system and the computational requirement due to the huge information supported by the vision systems compared to poor sensors like range finder. Furthermore, detection and tracking of the road boundaries is a major problem to achieve autonomy of vehicles that work in hazardous environment. This dissertation adopted a method based on lane marker recognition navigation with more merits than other visual navigation methods, such as the setting up and changing of mark line is relatively easy, low technical cost and so on. But the primary problems are the image distortion rectifying and border detection. This dissertation, taking car-like of two driving wheels as the research object, proposes a simple algorithm—calibration, to rectify the lane marker image. The field experiment result shows that the image-rectifying algorithm is efficient and provides the correct information for the recognition of white lane marker. A method about image processing is put forward with simple principle including the improved image segmentation of optimal threshold, the little data-processing filter based on mathematical morphology and border detection based on two-stage derivative. Meanwhile, the way to get the relative position of the body is improved. Fuzzy-optimal controller is introduced in the simulation. The results of simulation show that the control system can make the intelligent vehicle get good performance on control effects. In order to testify the correctness of the above image processing algorithm and tracking controller and the working effect of the intelligent vehicle, the main procedure (including dynamic image processing, controller and data communication) on the upper-stage computer and motor-controlled procedure on the lower-stage chip are compiled and the tracking path experiment is implemented on an intelligent vehicle experimental platform with visual navigation. The experiment results show that the procedure based on the above algorithm can perform the real-time image-processing task, and the intelligent vehicle has the strong capability against noise disturbance and can carry out the path tracking accurately and reliably.更多还原