摘要
针对传统LK光流法无法有效跟踪快速移动的目标,跟踪时存在特征点选取实时性和准确性不足的问题,提出一种结合改进FAST角点检测的LK光流算法。应用改进后的FAST角点检测提取出的角点作为候选特征点,通过设定筛选方案从中选取具有较高对比度的特征点作为目标特征点,结合图像金字塔分层,最终使用LK光流法对其进行跟踪。改进FAST角点检测能够更快速地提取出最强的灰度变化角点,使得子像素计算准确性得以提高并且减少了提取特征点的时间。引入图像分层缩放源图像,能够使算法稳定跟踪快速运动的目标。实验结果从运动目标检测所需时间、特征点的数量、每秒处理的视频帧数以及x轴和y轴方向运动误差等方面进行分析比较,证明所提出的改进算法运行速度快,能够快速且准确地跟踪动态目标。
The traditional LK(Lucas⁃Kanade)optical flow method fails to track fast⁃moving targets.In addition,its real⁃time performance and accuracy of the feature point selection is unsatisfactory in the process of tracking.Therefore,an LK optical flow algorithm based on the improved FAST corner point detection is proposed.The corner points extracted by the improved FAST corner point detection are used as the candidate feature points.The feature points with high contrast are selected as the target feature points by setting a screening scheme.In combination with the pyramidal image layering,the LK optical flow algorithm is used to track the targets,finally.The improved FAST corner point detection can extract the strongest grayscale variation corner points more quickly,which improves the accuracy of sub⁃pixel calculation and reduces the time for extracting feature points.The introduction of image hierarchical scaling source images can make the algorithm track fast⁃moving targets stably.The experimental results,including the time required for motion target detection,the number of feature points and the number of video frames processed per second,are analyzed and contrasted,as well as the motion error in the directions of x⁃axis and y⁃axis.It has been proved that the proposed improved algorithm has fast operation speed and can track the dynamic targets rapidly and accurately.
作者
朱代先
刁弘伟
刘树林
ZHU Daixian;DIAO Hongwei;LIU Shulin(College of Communication and Information Engineering,Xi’an University of Science and Technology,Xi’an 710054,China;College of Electrical and Control Engineering,Xi’an University of Science and Technology,Xi’an 710054,China)
出处
《现代电子技术》
2022年第15期45-50,共6页
Modern Electronics Technique
基金
陕西省重点研发计划(2021GY⁃338)。
关键词
LK光流法
FAST角点检测
目标跟踪
动态目标
金字塔图像分层
迭代
特征点
LK optical flow algorithm
FAST corner point detection
target tracking
dynamic target
pyramidal image layering
iteration
feature point
