融合多源图像与光流运动的旋转背景下对地运动目标检测研究

项目名称： 融合多源图像与光流运动的旋转背景下对地运动目标检测研究

项目编号： No.61471194

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 无线电电子学、电信技术

项目作者： 闫钧华

作者单位： 南京航空航天大学

项目金额： 83万元

中文摘要： 对地运动目标的高精检测是战机实现精确打击的关键问题。目前国内外的研究成果主要针对战机平飞或大范围盘旋时的运动目标检测，当战机急剧转弯导致图像地面背景剧烈旋转运动时，会产生大量漏检、虚警。鉴于此，本项目首次提出融合多源图像与光流运动的旋转背景下对地运动目标高精检测方法及半物理实验验证。建立数学模型描述战机、光电探测平台、摄像机的空间运动与姿态变化多因素联合影响下的图像地面背景与目标的运动特性，提出地面背景旋转运动剧烈程度的定量衡量指标，确立地面旋转运动背景中运动目标检测的理论依据。为综合利用可见光与红外图像及其图像中的运动信息，在配准可见光与红外图像基础上，融合红、绿、蓝、红外热辐射四维信息和光流运动，在图像高维融合光流场计算过程中实现多源图像信息融合。构建包括整体地面背景旋转运动与局部目标运动信息的图像运动场，从中重构纯地面背景旋转运动场，据此分离地面旋转背景与运动目标，实现运动目标检测。

中文关键词： 目标检测；运动识别；背景重构；视频图像

英文摘要： Ground moving target detection with high precision is the key issue to achieve precision strike of aircrafts. Current ground moving target detection algorithms of domestic and overseas are mainly for the situation in which the aircraft is in level flight or wide range of circling. When the aircraft makes sharp turns, which causes intense rotational motion of the background in the video, those algorithms will produce a large number of undetected or false alarms. In view of this, our project first proposes a high-precision moving target detection method based on the integration of multi-source images and optical flow motion and prepares semi-physical detection experiments. We establish the mathematical model to describe the motion characteristic of image ground background and target with the combined impact of spatial movement and attitude change of the aircrafts, photoelectric detection platform, and the camera. We propose a quantitative indicator to measure the severity of background ground rotational motion, establish moving target detection theory in background ground rotational motion. Then, to use visible and infrared images and image motion information, we merge four dimensional of red, green, blue, infrared thermal radiation information and optical flow motion on the basis of registration of visible and infrared images, achieving multi-source image information fusion in the calculation of high-dimensional image optical flow field. Ultimately, we construct image motion field which includes the information of whole background rotation movement and targets motion, reconstruct the pure rotating motion field of ground background from it, separate rotating ground background and moving targets, achieve moving target detection.

英文关键词： target detection;motion recognition;background reconstruction;pictures of video