摘要
针对传统方法目标装饰图像特征点采集效率低且图像匹配准确率低的问题,提出一种目标装饰图像多维特征点集自适应跟踪方法。结合SIFI算法确定图像极值点,去除不稳定的边缘响应点,以极值点为中心,结合不同尺度的高斯核函数和图像卷积,得到尺度空间中的装饰图像,以欧式距离作为匹配相似性度量值,完成对装饰图像的特征向量匹配,从而生成匹配特征点集集合。采用光流跟踪法确定装饰图像的可靠性特征点集,利用仿射变换参数剔除掉被误处理的可靠特征点,结合显著性特征确定目标装饰图像跟踪框,实现对目标装饰图像多维特征点集的自适应跟踪。仿真结果表明,所提方法跟踪效率高达96%,图像特征匹配率高达95%,说明该方法能够高效准确地采集目标装饰图像特征。
Aiming at low efficiency and accuracy of image matching in traditional methods,it proposes a multi-dimensional feature point set adaptive tracking method for target decoration image.Combining with the SIFI algorithm to determine the image extremum,it removes the unstable edge response point.Taking the extremum as the center,combining with Gaussian kernel function and image convolution of different scales,it obtains the decorative image in the scale space.Using the Euclidean distance as the matching similarity measure,it realizes the feature vector to match the decorative image,and generates the matching feature point pair set.Based on optical flow tracking method it determines the reliability feature point set of decorative image.Combining the salient features,it realizes the adaptive tracking of the multi-dimensional feature point set of the target decoration image,determines the reliable feature points and the target decorative image tracking frame The simulation results show that the tracking efficiency of the proposed method is as high as 96 and the matching rate of image features is up to 95.
作者
李瑾
Li Jin(Department of Art,Shaanxi Preschool Normal University,Shaanxi Xi'an,710010,China)
出处
《机械设计与制造工程》
2019年第7期96-99,共4页
Machine Design and Manufacturing Engineering
关键词
图像特征点
SIFT算法
仿射变换参数
自适应跟踪
image feature point
SIFT algorithm
affine transformation parameters
adaptive tracking
