基于大小场景整合的遥感小目标检测算法

Small target detection algorithm of remote sensing data based on LF-SF integration

针对目前遥感图像小目标检测算法遇到的复杂背景建模困难、先验信息严重匮乏等问题,考虑到昆虫视觉系统在图像处理方面的优势,提出基于蝇视觉系统大小场景(LF-SF)整合的信息处理模式解决遥感图像存在复杂背景的目标检测。蝇视觉的LF-SF整合机理无需考虑图像背景的复杂度以及目标先验信息,在抑制复杂背景纹理特征的同时对稀有目标特征进行增强,最后通过加权融合实现目标检测。实验结果表明,本文算法在算法设计、处理速度以及检测精度均优于现有算法。

Aiming at the present problems of small target detection algorithm for remote sensing data, such as hard to model background feature and lack of prior information seriously,considering the advantages of insect vision on image processing, this paper proposes a parallel processing model to solve the problem of targei detection of remote sensing under clutter background, which is inspired by the large field and small field （LF-SF） integration theory in fly vision system. Without background model and pri- or information, the theory of LF-SF integration can reduce the influence of background feature, enhance ＂the seldom target feature at the same time,and at last it uses weighted fusion to complete target detection. Experiment results show that the proposed method is better than the present algorithms in compu tation quantity,speed and accuracy of detection result.