基于预测的无人机视频快速匹配算法研究  

Fast Matching Methods for UAV Video with Prediction

作　　者：杨德龙[1] 吴云东[2] 蔡国榕[2] 陈水利[3] YANG De-long WU Yun-dong CAI Guo-rong CHEN Shui-li(School of Science, Jimei University, Xiamen 361021, China School of Computer Engineering, Jimei University, Xiamen 361021,China Chengyi University College, Jimei University, Xiamen 361021,China)

机构地区：[1]集美大学理学院,福建厦门361021 [2]集美大学计算机工程学院,福建厦门361021 [3]集美大学诚毅学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2016年第5期370-378,共9页Journal of Jimei University：Natural Science

基　　金：国家科技支撑项目(2013BAC08B01-01)

摘　　要：针对无人机视频数据,提出一种快速逐帧匹配算法。该算法首先使用计算机在每一帧中均匀标记一定数量的点,接着利用视频相邻帧之间的特点使用绝对差值法完成同名点的寻找,之后引入内插理论建立预测模型,缩小了搜索窗口和匹配窗口大小,降低了算法时间复杂度。实验结果表明,该算法可以高效地完成相邻帧之间的匹配。This paper proposes a fast matching algorithm for UAV^s video. Firstly, the algorithm use computer to sign some points which are uniform distribution on the image. Secondly, the algorithm take ad-vantage of the adjacent frame,s characteristics to find the correspond points. At last, we introducing the inter-polation theory to establish the prediction model to reduce the size of the searching window and matching win-dow ,make the time complexity lower than before.

关 键 词：相邻帧 模板匹配 PID 预测 内插 

分 类 号：N941.5[自然科学总论—系统科学]