摘要
图像光流的计算不需要在图像序列中建立特征之间的对应关系,但是直接利用光流法进行目标的运动参数估计和跟踪运算量很大,而且受噪声等影响而比较困难。首先采用聚类分割方法提取出目标区域,然后对目标区计算光流场,得到目标运动速度估计值,实现目标跟踪。此方法对比较复杂的红外图像也可以进行较好的分割,跟踪时无需建立图像间的特征对应关系,计算量大大缩小,目标运动速度的估计误差也很小。
In the process of computing optical flow, the corresponding relation between features of sequent images isn't required. However, to estimate motion parameter by the optical flow field directly is difficult because of noise and large amounts of calculation. A new clustering segmentation method is introduced to extract the target region at first, then optical flow of the target is calculated, and the speed of target is estimated at last. This method could extract the infrared target in complex background, the time shortens, and the error of estimation is also small.
出处
《光学技术》
CAS
CSCD
2003年第1期63-65,共3页
Optical Technique
