摘要
针对遥感图像在光照和几何差异等复杂因素上引起的匹配误差,深入研究了SIFT特征描述符的改进方法。利用特征点圆形区域来构造特征描述符,采用自适应量化策略用来局部区域的划分和梯度直方图的计算,并对每个描述子采用一种插值法重新确立主方向,改进SIFT算法的描述符。降低特征点维数的同时,又保证特征点描述符的独特性和鲁棒性。研究结果表明,改进的SIFT描述符在遥感图像几种复杂环境下都取得了预期的结果,证实了改进算法的可行性。
For remote sensing image matching error on the complex factors causing differences in light and geometry, in-depth study of the improved method of SIFT feature descriptor. Use a circular area to construct the feature point feature descriptor, adaptive quantization strategy used to calculate the histogram division and gradients local area, and each deseriptor uses an interpolation method to re-establish the main direction, improved SIFT algorithm descriptor. Reduce the dimension of feature points, they also feature point descriptor ensure uniqueness and robustness. The results show that the improved SIFT deseriptors in the complex envi- ronment of several remote sensing images have achieved the expected results, confirming the feasibility of the improved algorithm.
出处
《电视技术》
北大核心
2016年第9期108-111,共4页
Video Engineering
基金
重庆市研究生科研创新基金项目(CYS15166)
关键词
SIFT算法
特征匹配
特征描述符
主方向
自适应分级
遥感图像
SIFT algorithm
feature matching
feature descriptor
main direction
adaptive classifieation
remote sensing image