基于语义特征的视觉定位研究

Research on Visual Location Based on Semantic Characteristics

传统的视觉定位是通过提取图像的低级几何特征进行匹配从而实现对自身位姿的确定,这种特征不仅容易受角度、光照等的影响,而且缺少对场景的理解,不能利用常见导航地图,不易与人进行交互。为此提出了一种采用语义特征的视觉定位,利用视觉传感器及语义地图,通过语义特征的提取和匹配,实现对所处环境的感知并明确自身的位置。这种视觉定位能够弥补低级特征的缺点,并且能有效的利用现有地图,对周围环境进行感知。

The traditional visual location is to determine the pose by matching the low-level geometric features of the image.This feature is not only easily affected by angle and lighting,but also lacks understanding of the scene and can’t use common navigation maps.It is not easy to interact with people.This paper introduces a way of visual positioning based on semantic features,using visual sensors and semantic maps,and realizing the perception of the environment and location through extracting and matching semantic characteristics.This visual positioning can make up for the shortcomings of low-level features,and effectively utilize existing maps to perceive the surrounding environment.