结合显著性检测与特征匹配的飞机目标识别  

Aircraft target recognition based on saliency detection and feature matching

作　　者：欧阳欢 范大昭[1] 郭静 代亚贞 OUYANG Huan;FAN Dazhao;GUO Jing;DAI Yazhen(Information Engineering University,Zhengzhou 450001,China;Troops 93920,Hanzhong 723213,China;Troops 61206,Beijing 100042,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]93920部队,陕西汉中723213 [3]61206部队,北京100042

出　　处：《测绘通报》2020年第3期1-6,共6页Bulletin of Surveying and Mapping

基　　金：国家自然科学基金(41401534)。

摘　　要：根据目标与背景间的视觉显著性差异及飞机目标的形态特征,本文提出了一种基于显著性检测的飞机目标识别算法。该算法首先对预处理后的遥感影像进行显著性检测并提取场景中的感兴趣目标;然后依据阴影在HIS颜色空间的特性消除阴影对飞机识别造成的影响,并以飞机的形态特征为依据对感兴趣目标进行筛选,进而实现飞机目标识别;最后,计算飞机的外形参数、相对矩等特征并以此作为依据实现飞机机型判别。试验证明,该方法飞机特征提取精确,目标识别准确率高,具有较好的稳健性。According to the difference of visual saliency between target and background together with the morphological characteristics of the aircrafts,this paper proposes a novel aircraft target recognition algorithm based on saliency detection. The algorithm first detects the saliency of the pre-processed remote sensing image and eliminates the influence of shadow for the region of interest extraction.Distinguish the region of interest based on the morphological characteristics of the aircrafts,and then achieve aircraft target recognition.Finally,calculates the aircraft fuselage length,wingspan and relative moments. Afterwards,the aircraft type discrimination can be achieved via feature matching. Extensive experiments show that the algorithm has accurate feature extraction results and high target recognition accuracy.

关 键 词：遥感影像 飞机识别 显著性检测 特征匹配 机型判别 

分 类 号：P23[天文地球—摄影测量与遥感]