基于特征点光流聚类的复杂背景中运动车辆检测

为了准确、快速的在动态场景中对运动车辆进行检测,提出一种基于特征点光流聚类的车辆检测方法。该方法取Harris角点为特征量,通过对特征点做光流提取来剔除一些没有运动的干扰角点,然后再通过模糊U邻域(FUNN)聚类算法剔除噪音、孤立点和不感兴趣样本并实现前景和背景的分离,最后通过设定阈值判断前景目标是否是车辆。实验结果证明在复杂的动态场景中该算法具有更高的车辆识别率。

Moving Vehicle Detection in Complicated Background Based on Optical Flow of Feature Points

In order to detecting the moving vehicles accurately and fast in dynamic scene. This paper proposes a vehicle detection method which based on optical flow clustering of feature points. Harris corner points are chosen to form a feature vector in this method. We remove the interference corner points that are not moving based on extracting the optical flow of the feature points. Then we remove the acnode noise and non-interesting samples and realize the separation of foreground and background by the Fuzzy U Nearest neighbor Adaptive Clustering Algorithm. Finally, we determine whether the foreground is a vehicle by setting the threshold. The experimental results show that this algorithm has a high recognition rate to vehicles in complex dynamic scene.