杨燕,胡小鹏,吴思宁,王凡,刘培启.基于注意力选择的局部特征匹配方法[J].,2019,59(2):186-193 |
基于注意力选择的局部特征匹配方法 |
Matching method of local feature based on attention selection |
|
DOI:10.7511/dllgxb201902011 |
中文关键词:特征匹配注意力选择重复结构三角约束 |
英文关键词:feature matchingattention selectionrepetitive structuretriangle constraint |
基金项目:辽宁省教育厅科技研究资助项目(L201683682);国家自然科学基金资助项目(21676048). |
|
摘要点击次数:678 |
全文下载次数:632 |
中文摘要: |
为提高局部特征匹配性能,提出了基于人类注意力选择机制的特征匹配方法.通过局部特征的显著度分析,优先选择区域显著的特征进行匹配,利用可靠的匹配点对进行三角区域约束,实现区域内重复特征的精确匹配.实验结果表明,通过逐步缩小对应匹配搜索区域的方法抑制由相似特征引起的匹配歧义性,兼顾了高匹配精度和查全率. |
英文摘要: |
To improve the matching performance of local features, a feature matching method based on human attention selection mechanism is proposed. Based on the saliency analysis for local features, to achieve the accurate matching of repetitive features in region, the most salient local features are prefered to be matched and the robust correspondences are found to perform the following triangle constraint. The experimental results show that the reduction of the matching region is effective for reducing the matching ambiguities of local feature. The matching performance including both the matching precision and the recall rate is improved, especially for the scenes with repetitive structures. |
查看全文查看/发表评论下载PDF阅读器 |
| --> 关闭 |