Research on vanishing point detection of unstructured road scene combined with stereo vision
Q3 Engineering
引用次数: 0
Abstract
消失点检测是基于视觉的无人车辆自主导航的重要组成部分。由于非结构化场景存在缺乏清晰的道路线和复杂的背景干扰等问题, 现有检测方法普遍存在精度低、计算时间长的缺点。因此, 针对非结构化道路特点, 提出了一种结合立体视觉的消失点检测方法。采用双目立体视觉技术获得道路图像的视差图, 使用广度优先算法快速估计出道路图像的背景区域; 设计四方向五尺度的Gabor滤波器组估计像素响应幅值, 并通过幅值校正减少检测误差; 结合背景区域设计一系列投票点选择策略, 来剔除背景区域的干扰, 提高算法精度; 采用动态调整候选点范围策略, 减少消失点的搜索范围, 从而提高算法效率; 设计了一种角度优先投票函数, 将在投票空间中获得最大票数的候选点视为消失点。结果表明, 改进的方法在复杂背景干扰的场景下具有较好的鲁棒性, 在检测速度和检测精度上都有显著提升。结合立体视觉的非结构化道路场景消失点检测研究
Vanishing point detection is an important component of vision based autonomous navigation for unmanned vehicles. Due to the lack of clear road lines and complex background interference in unstructured scenes, existing detection methods generally have the drawbacks of low accuracy and long calculation time. Therefore, a vanishing point detection method combining stereo vision is proposed based on the characteristics of unstructured roads. Using binocular stereo vision technology to obtain disparity maps of road images, and using breadth first algorithm to quickly estimate the background area of the road image; Design a four sided five scale Gabor filter bank to estimate the pixel response amplitude, and reduce detection errors through amplitude correction; Design a series of voting point selection strategies based on the background area to eliminate interference in the background area and improve algorithm accuracy; Adopting a strategy of dynamically adjusting the range of candidate points to reduce the search range of vanishing points, thereby improving algorithm efficiency; We have designed an angle first voting function that treats candidate points that obtain the maximum number of votes in the voting space as vanishing points. The results show that the improved method has good robustness in complex background interference scenarios, and has significant improvements in detection speed and accuracy.
本文章由计算机程序翻译，如有差异，请以英文原文为准。
求助全文
约1分钟内获得全文
求助全文
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号
