关键词: 单目相机, 组合惯导, ＹＯＬＯｖ３, ＫＮＮ, 锥桶地图

Abstract: Aiming at competition items of straight-line acceleration,figure-eight circle and high-speed tracking of Formula Student Autonomous China (FSAC),this paper proposes a cone barrel map construction algorithm based on the fusion of monocular camera and combined inertial guidance.YOLOv3 is used to detect the type of the cone barrel,monocular camera ranging and combined inertial guidance positioning are used to determine the position of the cone barrel,and an improved K-nearest neighbors (KNN) algorithm is used to filter the obtained cone barrel.The experimental results show that,compared with Markov Random Field (MRF),the accuracy of the proposed algorithm increases by 4.3%,the recall rate increases by 2.9%,and the average error of the cone barrel position reduces by 52.8%,which can be used as a reference for the mapping of driverless sensing and positioning.The proposed algorithm also solves the problem of FSAC formula racing map construction well,and provides judgment basis for decision planning of FSAC formula racing.