基于单目视觉的区域交通智能车辆道路边界检测方法研究

基于单目视觉的道路边界检测由于其在车辆辅助驾驶系统中的重要应用价值成为当前计算机视觉和智能车辆领域最为活跃的研究课题之一。指出图像边缘检测现有算法的不足,采用领域平均法对图像进行平滑处理,根据图像的边缘特征运用Prewitt算子实现边缘增强,以获取精确的边缘信息。使用最大熵算法分割二值化图像进一步减少噪声,从而得到良好的道路特征图像数据。利用道路约束条件,建立视觉系统动态感兴趣区域(DAOI),运用改进的Hough变换最终识别道路边界。试验结果表明:本文所述算法不仅能准确、实时检测出道路板边界,而且能有效地抑制噪声,为区域交通智能车辆的换道和超车提供研究基础。

Study on Road Edge Recognition Method for Cybercar Based on Monocular Vision

Lane detection based monocular vision system has become one of the hottest topics in the domain of computer vision and intelligent vehicle because of its potential applications in driver assistance systems. Disadvantagesof traditional methods for image edge detection were discussed.Domain counterpoise algorithm was adopted to smooth the original image.According to edge feathers of the image,the image edge was enhanced by Prewitt algorithm to obtain exact edge information.The noise was further reduced after image segmentation by maximum entropy to obtain the image data about road feathers,Dynamic Area of Interest(DAOI)of vision system was established with road restrict condition and the road edge was recognized by using improved Hough transformation.The experiment result shows that the method can not only detect real-time lane edge precisely,but also restrain noise effectively.It provides a research foundation for lane changing and overtaking for JLUIV5-Cybercar in the field experiments.