基于引导滤波与视觉注意机制的红外小目标检测

发布时间：2018-09-13 12:38

【摘要】：将目标从背景中有效分割出来是红外运动目标检测与跟踪技术的一个关键技术。基于热成像原理的红外图像特点是噪声强、干扰多且目标边界模糊不清,尤其是当信号比较弱、热成像面积较小时,背景和噪声在图像中所占比例很大,从而导致对红外小目标的检测难度增大。因此,红外小目标检测技术的研究更具挑战意义。算法针对红外图像小目标的特点及其检测难点,采用引导滤波进行预处理,在保持目标边缘特性的基础上平滑背景增强背景之间关联性,进行降噪。利用视觉注意机制可以引导人眼在海量的数据中关注到对比度大的区域的特性,采用分层阈值化和连通域分析对图像进行感兴趣区域(Region of interest,ROI)提取,在感兴趣区域内采用多尺度LOG滤波提高目标的信杂比,通过信杂比阈值初步的到检测目标。为正确的检测真实小目标并过滤伪目标,对移动式管道滤波进行改进,充分利用管道中的已有目标的先验信息,实现对红外小目标的准确检测。本文对引导滤波的预处理效果和改进后移动式管道滤波进行了多目标、弱目标、背景复杂情况下的对比实验,实验证明其具有良好的有效性。同时对该算法进行多个视频的实验,计算其检测率与虚警率,结果表明该算法具有良好的适应性。
[Abstract]:The key technology of infrared moving target detection and tracking is to segment the target from the background effectively.The infrared image based on thermal imaging principle is characterized by strong noise, more interference and blurred target boundary.Especially when the signal is weak and the thermal imaging area is small, the background and noise occupy a large proportion in the image. Therefore, the research of infrared small target detection technology is more challenging. Aiming at the characteristics of infrared image small target and its detection difficulties, the algorithm uses guided filter to preprocess, smoothes the correlation between background and enhances the background on the basis of keeping the edge characteristics of the target, and denoises. Visual attention mechanism can guide the human eye to focus on the characteristics of the region with large contrast in a large amount of data. Region of interest (ROI) is extracted by hierarchical thresholding and connected area analysis, and the target's signal-to-clutter ratio is improved by multi-scale LOG filtering in the region of interest. In order to detect the real small target and filter the false target correctly, the mobile pipeline filter is improved to make full use of the prior information of the existing targets in the pipeline to realize the accurate detection of the infrared small target. The experimental results show that the algorithm is effective in the case of complex background and target. At the same time, the detection rate and false alarm rate of the algorithm are calculated by several video experiments. The results show that the algorithm has good adaptability.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】