摘要
针对传统的角点检测在复杂场景视频图像中角点检测的准确率不高、实时性差的问题,提出了基于ORB特征提取和GroupSAC的快速角点检测算法。首先在O-FAST中加入图像金字塔对图像进行角点检测,保证了视频图像角点检测的尺度不变性,然后利用灰度质心对求得的主方向进行BRIEF特征提取,保证了视频角点检测的旋转不变性。最后利用基于混合二项式模型的GroupSAC算法去除误差点,提高了角点检测的准确率。实验证明,在复杂场景视频图像中,如在光照变化、尺度变化、旋转变化以及遮挡图像的情况下,该方法具有较高的准确性、较低的时间复杂度和较强的鲁棒性,实现了对复杂场景视频图像的快速角点检测。
Aiming at the problems of corner detection accuracy is low and real-time performance is poor in complex scenes in video image by the traditional corner detection,the fast corner-point detected algorithm based on orb feature extraction and Group SAC algorithm is presented. Firstly,to ensure that the algorithm has a scale invariant features,O-FAST algorithm that adding image pyramid is adopted to detect corner. Then extract features by BRIEF that using the gray centroid method obtains the main direction to have better rotation invariance. Finally,the mixed binomial model Group SAC algorithm removes errors in order to ensure the accuracy of the corner. And it is applied in video images of complex scenes,such as changes in illumination,scale change,rotation changes and occluded image. The experimental results show that the algorithm improves the accuracy,reduces the time complexity and has strong robustness.
作者
王丽芳
赵雅楠
秦品乐
高媛
WANG Li-fang ZHAO Ya-nan QIN Pin-le GAO Yuan(School of Computer Science and Control Engineering, North University of China, Taiyuan 030051 ,P. R. China)
出处
《科学技术与工程》
北大核心
2017年第2期88-94,共7页
Science Technology and Engineering
基金
山西省自然科学基金(2015011045)资助
