| 摘 要: 针对传统光流法图像特征缺失、边界及遮挡等处容易导致目标跟踪质量降低或丢失的问题，提出一种半稠密光流法实现图像特征的稳定跟踪。首先计算出图像中像素梯度变化较大的像素区域；其次利用时变图像灰度的时空梯度函数来计算像素的速度矢量，进而实现像素区域的跟踪；最后将状态向量作为剔除跟踪失败的依据，保留跟踪质量优良的像素区域。结果表明，该算法能有效地提高图像特征跟踪能力，保证图像中重要信息不丢失，同时保证运算速率。 |
| 关键词: 光流法；半稠密；图像处理；特征跟踪 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 重庆市教委科学技术研究项目(KJQN201905603，KJQN201901123)，重庆市科技局技术创新与应用发展重点项目(cstc2019jscx-mbdxX0002). |
|
| Image Tracking and Matching Algorithm of Semi-dense Optical Flow Method |
|
YU Xiaoyi1，SONG Tao2，WEI Paifeng1，MA Aiping1
|
( 1.Chongqing Energy College, Chongqing 402260, China ;
2.Chongqing University of Technology, Chongqing 400054, China)
495527898@qq.com; tsong@cqut.edu.cn; st20052862@163.com; 2639251078@qq.com
|
| Abstract: Traditional optical flow method tends to cause poor quality or loss of target tracking in areas of image feature missing, boundary and occlusion. In view of this problem, this paper proposes a semi-dense optical flow method to achieve stable tracking of image features. First, pixel area with great gradient changes in the image is calculated. Secondly, pixel velocity vector is calculated by spatiotemporal gradient function of time-varying image gray in order to realize pixel area tracking. Finally, state vector is used as a basis for removing tracking failure and pixel area with good tracking quality is retained. Results show that the proposed algorithm can effectively improve image feature tracking ability and ensure important information in the image and calculation rate. |
| Keywords: optical flow method; semi-dense; image processing; feature tracking |
