LIU Tian, ZHANG Zhengyu, WANG Xueyuan, et al. Real-time recognition method of image marker points based on motion estimation[J]. 2020, (8): 65-67+71. DOI: 10.13873/J.1000-9787(2020)08-0065-03.
基于运动估计的图像标记点实时识别方法
摘要
针对高分辨率视频测量(VM)中标记点难以实时识别的问题
提出基于运动估计的图像标记点实时识别方法
通过标记点运动估计模型获得标记点在时序图像中的预估坐标
根据各预估坐标对图像并行分割获得对应搜索域
并采用增量更新的Freeman链码提取标记点轮廓信息
从而达到实时识别标记点的目的。运动估计实验表明:该模型能实现风洞环境中图像标记点运动估计
最大误差仅为5.597像素;标记点并行识别实验表明:利用现场可编程门阵列(FPGA)实时识别标记点
处理效率达到2GB/s
应用前景广阔。
Abstract
Aiming at the problem that the marking points in high-resolution video measurement( VM) are difficult to identify in real-time
a real-time recognition method for image marking point based on motion estimation is proposed. The estimated coordinates of the marker points in the time series image are obtained by the marker point motion estimation model. The corresponding search domain is obtained by parallel segmentation of the images according to the estimated coordinates
and the corresponding search domain is obtained. The incremental updated Freeman chain code extracts the contour information of the marker points
thereby achieve the purpose of identifying the marker points in real-time. Motion estimation experiment shows that the model can estimate the motion of image marker points in wind tunnel environment with a maximum error of 5. 579 pixels. Parallel recognition experiment shows that using FPGA to realize marker point recognition
the real-time processing efficiency reaches 2 GB/s