知識庫

Implementation of Dynamic Boundary on Multiple Kalman Tracking using Radar (利用動態區間設定多障礙物的雷達卡爾曼追蹤)
  • 發布年度:2017
  • 主要類別:電動車與車輛電子
  • 次要類別:論文
  • 利用卡爾曼濾波器追蹤多重障礙物,此論文根據歷史追蹤紀錄決定資料點消失補償,以預測值取代觀測值並適時中斷不好的追蹤點。在多重追蹤下,雷達不同於影像,無法判定主要障礙物,此論文以動態區間設定方式,找出危險性主障礙物。此技術可應用於多重危險主障礙物追蹤。

    Abstract—The development of advanced driver assistance systems (ADAS) is of importance to prevent the potential driving risks on the road, which relies on the perceptive sensors to detect and identify the traffic conditions. This paper examines a method to track multiple objects and determine the high risk obstacles using radar. We experiment the scenarios as the vehicle and pedestrians in front, as well as the pedestrian crossing the street. The multiple tracks are achieved based on the Kalman filter. This work implements the state prediction model to compensate the missing targets, and the false positives are excluded through a threshold of bad tracks. To identify the potential obstacles, we impose two boundaries, i.e., lane and dynamic boundaries as the region of interest on the multiple tracks. It is demonstrated that the dynamic boundary benefits the finding of high risk bstacles and excludes most of the background noises. The lane boundary takes the advantage of tracking bjects along the driving path: however, it could underestimate the risk of transverse moving objects in which the tracks are not continuous crossing the lanes.