以反透視映射模型與移動補償為基礎的障礙物偵測應用在倒車行為的安全上

Abstract

近年來，交通事故造成人員受傷或死亡的數量仍然居高不下，因此世界各地提升安全性的高效能車輛與日俱增，其中車輛與其他障礙物碰撞事故發生最為頻繁，很多研究人員提出利用立體視覺的一般障礙物偵測技術，然而，影像式駕駛輔助安全系統中在移動車輛中以單支攝影機的障礙物偵測仍然是一個困難且具有挑戰性的問題。在本篇論文中，我們提出一個以移動補償與反透視映射模型為基礎的單眼影像式障礙物偵測技術，提出一個創新的地面移動量估測技術藉由利用路面偵測取得影像中最有用的地面特徵點，並且分析特徵點的光流主要分布，可以準確的獲得地面移動補償量，此外，利用反透視映射模型對影像中平面與非平面物體的不同特性來偵測障礙物。為了要得到更可靠的偵測結果與最靠近的可能碰撞位置，一個垂直方向的統計將被用來定位物體。我們提出的方法將會估計目標物離本車的距離來做更進一步的安全應用，最後我們測試了很多在日常生活倒車場景中可能會發生的情況，藉由實驗結果可證明我們提出的系統在障礙物偵測與距離估測上可達到一定的穩健性與準確性。 本論文的目標是希望藉由偵測阻礙行進路線上或突出地面具有一定高度的物體來提升倒車行為的安全，此外，藉由標示目標物離本車最近的位置與距離，可有效的避免在倒車上的碰撞。

The number of road traffic accidents which involves the fatality and injury accidents is so much in recent years, so that the number of higher performance vehicles for safety is growing up over the world. Among of all the collision accidents between vehicle and other generalized obstacles occur most frequently. For generic obstacle detection, researchers have proposed many methods which focus on stereo vision. However, automatic obstacle detection with a single camera system mounted on a moving vehicle is still a challenging problem in vision-based safety driving support systems. In this paper, we present a monocular vision-based obstacle detection algorithm which is based on movement compensation and IPM. We proposed a novel ground movement estimation technique which is employing road detection to assist in obtaining most useful ground features in the image, and analysis the principal distribution of optical flow of these feature points, the ground movement for compensation is obtained accurately. Besides, adopting different characteristics between planar and non-planar object result from IPM to detect obstacle. In order to get more reliable detection and the nearest collision position, a vertical profile is used to locate the obstacle. Furthermore the proposed technique is to measure the distance of target how far away ego-vehicle for safety application. This system has been tested for many conditions which could occur when backing up in real life situations, and the experimental results already demonstrated the effectiveness and accurateness of the proposed obstacle detection and distance measurement technique. The objective of this paper is to improve the safety for backing up maneuver by detecting objects that can obstruct the vehicle’s driving path or anything raise out significantly from the road surface. Besides, backing up collision can be avoided effectively by marking the nearest position and indicating the distance away from our ego-vehicle.