学术论文

      惯性信息辅助的大视角目标快速精确定位

      Fast and accurate target positioning with large viewpoint based on inertial navigation system information

      摘要:
      目标定位技术广泛应用于航空领域的侦察机、无人机等各类侦察打击任务中,目标定位精度的高低及效率对作战效果具有重要影响.针对仿射尺度不变的特征变换(ASIFT)算法对远距离大视角目标定位精度较低、速度较慢的问题,提出了一种基于惯性信息辅助的大视角目标快速精确定位方法.该方法首先对目标实测序列图像构造尺度空间,结合FAST特征检测与FREAK特征描述的方式进行匹配,实现对待定位目标的快速提取;然后利用机载惯性信息求解目标实测图与参考图之间的透视变换矩阵,利用该矩阵对实测图进行变换以减小图像间视角差异,克服了ASIFT算法盲目匹配计算的弊端,并通过FAST特征检测与FREAK特征描述相结合的方式提升了大视角图像的匹配速度;最后通过单应性矩阵映射关系实现对目标的精确定位.实验结果表明,大视角目标快速精确定位方法匹配耗时比ASIFT算法的减小了1个数量级,定位精度比目标平均值定位算法精度提高了1个数量级,有效提高了图像匹配定位在航空领域的应用效率.
      Abstract:
      Target location technology is widely used in aircraft and missiles for getting the position information of air-attack targets.The attack effect is directly affected by the efficiency of the target localization method.Considering the problem of low precision and slow speed of the Affine Scale-lnvariant Feature Transform (ASIFT) algorithm,a new positioning method based on inertial navigation system information is proposed.The scale space of the target real-time sequence images is constructed,and is matched with the combination of Features from Accelerated Segment Test (FAST) and Fast REtinA Keypoint (FREAK) description.Fast extraction of the target is then achieved.The homography matrix between the real-time and the reference images is solved by using the information of the airborne inertial navigation system.The real-time images are then transformed by the homography matrix to reduce the difference between the real-time and the reference images,so as to overcome the problem of blind multiple matching calculation of the ASIFT algorithm.Through the combination of FAST feature detection and FREAK feature description,the matching speed of large viewpoint images is enhanced.The target is accurate ly located in the reference image by the homography matrix.The experimental results indicate that the calculation speed is increased by one order of magnitude compared with that of the ASIFT algorithm,and the positioning accuracy is increased by one order of magnitude compared with that of the existing target average positioning algorithm.The method proposed in the paper will be helpful in improving the efficiency of the figure-matching in the aerial application.
      作者: 曾庆化 [1] 潘鹏举 [1] 刘建业 [1] 王云舒 [1] 刘昇 [2]
      Author: ZENG Qinghua [1] PAN Pengju [1] LIU Jianye [1] WANG Yunshu [1] LIU Sheng [2]
      作者单位: 南京航空航天大学自动化学院,南京 211106;卫星通信与导航协同创新中心,南京 211106 中航工业洛阳电光设备研究所,洛阳,471009
      刊 名: 航空学报 ISTICEIPKU
      年,卷(期): 2017, 38(8)
      分类号: V243
      在线出版日期: 2017年9月29日
      基金项目: 国家自然科学基金,江苏省普通高校研究生科研创新计划项目,中央高校基本科研业务费专项资金(NS2015037) National Natural Science Foundation of China,Funding of Jiangsu Innovation Program for Graduate Education,the Fundamental Research Funds for Central Universities