一种面向智能车辆视觉系统的电子稳像算法

An Electronic Image Stabilization Algorithm for Vision System of Intelligent Vehicles

摘要:

车载摄像头是智能车辆视觉系统中必不可少的部件。在恶劣道路或极限工况下，车辆的振动状况显著，车载摄像头采集到的图像序列会发生抖动。针对此问题，提出了一种适用于车辆复杂工况的电子稳像算法。基于车载工况下的实时性要求，选择ORB（oriented FAST and rotated BRIEF）算法进行特征检测与描述。为了提高特征点匹配精度与匹配效率，改进了传统随机采样一致性算法，增强了其对多匹配点、匹配点集中工况的适应性。为了适应车载工况下的剧烈振动，采用了自适应卡尔曼滤波算法以解决经典的卡尔曼滤波对初值敏感的问题。最后搭建了一辆振动特性显著的汽油模型车，在恶劣的路面条件下开展了实验，在较正常工况更为极端的条件下验证了提出的电子稳像算法的正确性与有效性。

关键词:电子稳像;视觉系统;智能车辆;自适应卡尔曼滤波;汽油模型车

Abstract:

Vehicular camera is an essential part of the vision system of intelligent vehicles. In harsh road or extreme conditions， due to the significant vibration of vehicles， the image sequence collected by the vehicular camera vibrates. Aimed at this problem， an electronic image stabilization algorithm for vehicle vision system is proposed. Considering the real-time requirements under vehicle conditions， the ORB algorithm （oriented fast and rotated BRIEF） is selected for feature detection and description. In order to improve the accuracy of matching and the efficiency of feature points， the traditional random sampling consistency algorithm is improved to enhance its adaptability to multiple and centralized matching points. The classical Kalman filter is sensitive to the initial value. Therefore， to adapt to the extreme conditions， the adaptive Kalman filter is used. Finally， a gasoline model vehicle with significant vibration characteristics is established， and experiments are conducted under harsh road conditions. The correctness and effectiveness of the proposed electronic image stabilization algorithm are verified under conditions more extreme than normal.

Key words:electronic image stabilization;vision system;intelligent vehicles;adaptive Kalman filter;gasoline model car