一种基于广义多尺度分形参数的小目标检测方法

Small Target Detection Based on Generalized Multiscale Fractal Parameter

复杂场景中小目标的检测是一个难点。本文从分数布朗随机场模型的描述出发，提出并构造了一种广义多尺度分形参数，用于描述图像的纹理信息。它可以作为区分分形集合与非分形集合的一种有效的测度。进一步，基于广义多尺度分形参数，本文提出了一种具有抗噪性能的稳健的目标提取算法，并用于海上小目标的提取。对于实际图像的实验结果表明，本算法具有极高的检测精度。

It is a difficult for the small target detection in complex scenes.In this paper,a kind of generalized multiscale fractal parameter is proposed and constructed for the description of image textures,which is based on fractional Brownian random field model.The generalized multiscale fractal parameter can be used as a effective measure for distinguishing the fractal sets and non fractal sets.Furthermore,an robust algorithm which is based on generalized multiscale fractal parameter is proposed for the small target extraction on ocean with good noise immunity.The experiments show that this algorithm has high detection precision for real images.This method can be used for reference for the small target detection in general complex natural scenes.