【摘要】 为了使车辆快速并准确地检测出人行横道,针对车载摄像头获取的路面图像,提出了一种基于决策树参数优化的人行横道检测方法,该方法分3个步骤完成:首先,通过逆透视变换获得指定区域内的道路俯瞰图;其次,通过阈值分割提取出路面白色标线;最后根据人行横道的特征属性实现检测。其中,白色标线的提取是准确检测的基础,因此结合边缘信息与自适应阈值,改进了分割算法,并使用决策树优化分割参数。结果表明:即使在路面光照不均匀的情况下,改进后的算法分割效果显示良好,有效提高了人行横道检测率。更多还原

【Abstract】 In order to make the vehicle quickly and accurately detect the pedestrian crossing, in view of the on-board camera for pavement images, a pedestrian detection method is put forward based on decision tree parameters optimization, This method works in three steps. First, the search for crosswalks is reduced to a suitable bird’s eye view of road surface by using inverse perspective mapping. Secondly, white lines are extracted through threshold segmentation. Lastly, crossing is detected according to its feature attribute. The second step is very essential for the following process. Hence, an improved algorithm is adopted based on local adaptive threshold and edge information, while optimization of segmentation parameters is achieved through decision tree. The experimental results show that the improved segmentation effect is good and the detection rate of crosswalk is improved effectively even when the illumination of the road surface is uneven.更多还原