基于机器视觉筛选GPS卫星信号的无人驾驶汽车组合导航方法

Integrated navigation system for unmanned intelligent vehicle based on vision

针对无人驾驶智能车在城市道路环境中全球定位系统信号以非视距传播而造成伪距估计偏差,导致定位不准确的问题,提出了一种以视觉信息为筛选条件的组合导航方法。该系统通过摄像头采集到车体周围环境信息图像并处理,最终得到建筑物上隅角与下隅角的差值,换算出建筑物相对车体所形成的遮挡角度,以此筛选GPS卫星有效信息,采用容积卡尔曼滤波对组合导航信息进行滤波估计,提高导航信息准确性。实验证明,该系统能有效弥补无人驾驶智能车城市道路环境中GPS信号偏差的缺点,较传统组合方式相比具有较高的可靠性及实用价值。

In view of the unmanned intelligent vehicles in urban road environment,the global positioning system（ GPS） signal to non-line-of-sight（ NLOS） propagation caused by the pseudorange deviation,lead to inaccurate positioning problems,puts forward a kind of method for integrated navigation.The system through the camera to process images,finally get the corner on the building and the angle difference of conversion of building block angle,which is formed by the relative body and screening of GPS satellite information effectively.The system adopts the adaptive cubature Kalman filtering for filter estimation,in order to improve the accu-racy of navigation information.Experiments show that the system can effectively recluce GPS signal deviation in urban road environ-ment,which has high reliability and practical value.