作者：马潇潇；于刚；李长春

第一作者单位：河南理工大学测绘与国土信息工程学院

摘要：无人机遥感在勘探救援、灾情评估、灾后重建等领域的重要性日渐凸显,而无人机影像数据处理的2大关键技术即快速拼接和信息提取仍处于研究阶段。针对SURF算法和SVM算法存在的问题,首先对SURF-64与SURF-36进行比较,并经实验证明采用64维向量描述特征点更适合于无人机影像数据处理,可达到高效、准确拼接的目的;然后采用2种改进SVM算法对拼接影像进行信息提取,并与传统SVM算法进行比较。实验表明,2种改进算法在信息提取精度、算法泛化能力方面均有不同程度的提高和增强。

Abstract：The unmaned aerial vehicle ( UAV) remote sensing system is becoming increasingly prominent in the field of exploration and rescue, disaster assessment and reconstruction etc. Untill now, two key technologies of UAV image data processing ( fast matching and information extraction) are still under research. In order to solve the problems of using SURF algorithm and SVM algorithm. The comparision of SURF-64 and SURF-36 is done. The experimental results show that choosing a 64 dimensional vectors to describe the feature points is more suitable for UAV image data processing. It has advantages of efficient, accurate and fast matching. Meanwhile, two kinds of improved SVM algorithms are employed to extract information from the matched-image and comparing with the traditional SVM algorithm. The experimental results show that the two improved algorithm can improve and enhance in the accuracy of information extraction and generalization ability.

基金：国家自然科学基金资助项目(61602512)；河南省基础与前沿技术研究项目(152300410098)；河南省教育厅科学技术研究重点项目(16A420006)；

关键词：SURF;SVM;无人机影像;数据处理算法;

DOI：10.16186/j.cnki.1673-9787.2017.06.011

分类号：TP751