摘要
为快速、准确地实现图像局部特征提取,提高图像中目标识别的实时性,提出一种基于FAST-SIFT组合的局部特征探测算法。使用FAST检测算法进行快速的特征提取,利用SIFT的128维描述子进行准确的特征描述,在匹配阶段采用基于SNR的朴素贝叶斯分类器,提高系统的鲁棒性。实验结果表明,与传统SIFT算法相比较,该算法可以更快速地实现局部特征提取,在杂乱背景中能准确识别出目标物体。
A local feature detection using FAST extraction combined with SIFT description was proposed to provide fast and accurate local feature extraction and achieve real-time object identification.FAST detection was applied for fast feature extraction,and a 128-dimensional SIFT descriptor was created for each extracted feature.To increase robustness and eliminate outliers in matching,signal to noise ratio(SNR)index that measured matched pairs' spatial consistency was introduced.Object identity was inferred by propagating SNR through a naive Bayes classifier.Experimental results demonstrate the performance and speed of the proposed method are superior to traditional feature-based approaches.
出处
《计算机工程与设计》
北大核心
2015年第10期2749-2753,共5页
Computer Engineering and Design
基金
国家科技支撑计划基金项目(2013BAH45F02)