基于SLIC分层分割的无人机图像极小目标检测方法

摘要

针对无人机可见光图像目标小、对比度弱的问题,本文提出一种基于简单线性迭代聚类(Simple linear iterative clustering,SLIC)分层分割的极小目标检测方法.首先使用预处理方法提高原始图像的对比度,并利用Top-hat融合方法进行初始分割以确定目标区域,其次利用SLIC方法完成目标精细分割,并采用改进的具有噪声的基于密度的聚类方法(Density-based spatial clustering of applications with noise,DBSCAN)对SLIC分割结果进行超像素聚类,最后提取目标的邻域熵等多种底层特征,使用特征匹配方式检测目标,获取最终检测结果.本文提出了一种全局检测和局部检测相结合的检测策略,极大提高了检测速度.仿真结果表明,本文方法可以有效提高无人机小目标的检测性能,加速检测速度.%For the problem of the small target and the weak contrast of UAV image,we propose a method for minimal target detection based on simple linear iterative clustering (SLIC) hierarchical segmentation.Firstly,pretreatment methods are utilized to improve the contrast of the original image,and Top-hat fusion is used as initial segmentation to detect the initial target area.Then SLIC segmentation method is utilized to obtain the fine segmentation,and improved density-based spatial clustering of applications with noise(DBSCAN) is introduced to accomplish ultra-pixel classification according to the segmentation result.Finally,the target is detected through feature matching by extracting the neighborhood entropy of the target and other low-level features.Also a detection strategy combining global detection and local detection is proposed to accelerate the detection speed.The experimental results show that the proposed method can improve the detection performance for the minimal targets in UAV image and accelerate the detection speed.