基于复合动态模型和证据融合架构的移动物体检测与跟踪方法

Moving object detection and tracking based on composite dynamic model and evidential fusion framework

针对现有方法中移动物体检测与跟踪的准确性精度较低的缺点,提出一种基于多传感器检测分类的移动物体描述和感知方法:建立了一个包含核心对象动态特征和分类描述的复合模型,在此基础上设计了一个基于证据框架的信息感知与融合方法,通过整合动态模型和不确定性特征来实现对移动物体的检测和跟踪。为了验证所提方法的有效性,在一辆安装有雷达、激光雷达和摄像头的演示车上进行了相关实验,在不同驾驶场景下针对行人、卡车和轿车三个移动物体进行了检测与跟踪,实验结果证明所提方法具有非常高的准确性。

In order to overcome the disadvantage of low precision of the accurate detection and tracking of moving objects in existing methods,this paper proposed a novel description and perception method based on multiple sensor detection and classification:the composite model containing dynamic property and classification description was established,based on this model,a perception and fusion method of information based on evidential framework was devised to realize detection and tracking of moving object by integrating dynamic model and uncertain feature.To validate the effectiveness of the proposed method,some experiments on a vehicle demonstrator with radar,lidar,and camera were carried out,the detecting and tracking of three objects including pedestrian,truck,and car under different driving scenarios was tested.The results show that the proposed method possesses high precision.