知識庫

A Novel Controller Design for Collision Avoidance Systems Using Sensor Fusion Method (多感知融合在車輛安全上之研究)
  • 發布年度:2014
  • 主要類別:電動車與車輛電子
  • 次要類別:論文
  • ABSTRACT


    With aim of creating collision avoidance system for high accurate object detection, this paper
    has investigated a sensor fusion method for proposed system by blending single camera and
    Millimeter Wave (MMW) radar. Our research uses MMW radar to detect front vehicles and
    get the position of vehicles that is relative to host vehicle, and use camera to recognize vehicle.
    The predicted trajectory can be instantaneously calculated according to the angle of steering
    wheel which is derived from yaw rate. Therefore, the high risky vehicle can be found in the
    trajectory, and proposed recognition method only needs to focus on the ROI which the area is
    selected from high risky vehicle. In this paper, a standard autonomous emergency braking
    (AEB) scenario which is car-to-car rear stationary (CCRs) has selected, the maximum speed
    of demonstrated vehicle is 30 km/h. The detection rate of this paper is over 95%, and the
    average error rate which compares between proposed method and experiment is about 2%.
    The proposed system that utilizes vision, MMW radar, and inertial sensors are implemented
    with theoretical application, and hardware integration, and furthermore the results show the
    ability of vehicle recognition with data fusion.


    Keywords: Sensor Fusion, Radar, Collision Avoidance, Recognition.

    本文將探討如何使用影像與毫米波雷達來進行多感知融合,並應用於前方防撞系統。這兩種感測器近十年來已被廣泛的用在各種不同的安全系統,然而,各種感測器本身都有各自的物理特性與限制,例如影像容易受天候的影響,而測距感測器則無法辨識物體的種類,基於這些問題,近年來多感知融合的技術開始被重視。本研究將以影像辨識車道線的輔助來找出前方高危險性的障礙物,當車道線辨識失效時,則使用陀螺儀感測器來量測車身姿態的搖擺率(Yaw-Rate),進一步計算出道路曲率與車身轉向角,再以路徑預測演算法來濾除不在車道上的物體,解決影像失效的問題,除此之外,影像透過毫米波雷達的融合,可快速找出影像中障礙物的所在位置,增加演算法的處理速度。本文提出的前方防撞系統分別建立在DSP與MCU的平台上,並裝設於實車上來進行實車驗證。
     
    關鍵詞:標線偵測、毫米波雷達,多感知融合、路徑預測。