基于多尺度特征提取的运动目标定位研究

从尺度不变特征变换算法的实时性与鲁棒性分析入手,针对其在特征点匹配精度和匹配效率两方面不足,提出一种基于运动目标的多尺度特征提取算法.该算法通过设计出一种基于尺度因子变化的高斯核模板尺寸自适应调整以及时间轴帧图像双向配准的办法,成功地修正了运动目标图像配准的实时性与鲁棒性的不足.实验证明,多尺度特征提取算法能够有效而快...

Object location technique for moving target based on multi-scale feature extraction

Began with the analysis of the real-time performance and robustness of SIFT, the paper optimized traditional algorithm in two aspects of matching time and accuracy. The article proposes a multi-scale feature extraction algorithm (MSFE) which is based on moving target. MSFE devised two improved ways. One way was presented that the size of template could automatically adjust along with the scale factor changing, the other is that a bidirectional feature points matching algorithm is proposed. By MSFE, the defect of the real-time performance and robustness is amended successfully while moving target images matching. Finally, experiment showed that by the use of MSFE algorithm, the sequential frame images based on time-axis could be matching quickly and accurately.