上海交通大学学报(自然版) ›› 2015, Vol. 49 ›› Issue (06): 876-883.

• 自动化技术、计算机技术 • 上一篇    下一篇

基于奇异值分解和叠加法的慢速小目标检测算法

高敬礼1,文成林2,刘妹琴1   

  1. (1.浙江大学 电气工程学院, 杭州 310027; 2.杭州电子科技大学 自动化学院, 杭州 310018)
  • 收稿日期:2015-03-15
  • 基金资助:

    国家自然科学基金项目(61273170,61174142,61271144,61273075)资助

Low-Speed Small Target Detection Based on SVD and Superposition

GAO Jingli1,WEN Chenglin2,LIU Meiqin1   

  1. (1. College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China; 2. College of Automation, Hangzhou Dianzi University, Hangzhou 310018, China)
  • Received:2015-03-15

摘要:

摘要:  在图像集同一像素位置噪声满足遍历性的条件下,分析了利用叠加法进行目标检测的可行性,在此基础上提出了一种基于叠加法的慢速小目标检测算法.首先,分析了目标不重叠、完全重叠和部分重叠时,目标合成图像和噪声合成图像的能量变化情况,分析表明随着叠加次数的增加,噪声合成图像的能量衰减快于目标合成图像,从而保证了合成图像中目标与噪声能量比的增加,为利用叠加法进行目标检测提供了可行性.其次,根据合成图像的奇异值特性,利用标准化奇异值的不均匀变化完成了目标检测.同时,仿真验证了目标合成图像和噪声合成图像的能量衰减情况,分析了叠加次数、目标尺寸与强度对目标可检测性的影响.

关键词: 奇异值分解, 叠加法, 目标检测

Abstract:

Abstract: In the case of the measurement noises obtained from the same pixel position in a set of images which are ergodic, the feasibility of superposition method for target detection was analyzed, based on which, a superposition-based lowspeed small target detection algorithm was proposed. First, the energy variation of the target synthesis images and the noise synthesis images were analysized, when the targets were non-overlapped, completely overlapped and partially overlapped. The analysis shows that with the increase in the superimposed number, the energy of the noise synthesis images is attenuated faster than that of the target synthesis images, thus ensuring an increase in the ratio of target to noise in synthesis images. Second, depending on the characteristics of singular values of the synthesis images, the uneven variations of the standardized singular values were used for target detection. The simulation verifies the attenuation of the target energy and noise energy of the synthesis images, and analyzes the impact of superimposed number, target size and strength on target detectability.

Key words: singular value decomposition (SVD), superposition, object detection

中图分类号: