SIA OpenIR  > 光电信息技术研究室
基于知识的前下视红外图像机场跑道识别算法研究
Alternative TitleRecognition of Airport Runway in FLIR Images Based on Knowledge
武伟1,2
Department光电信息技术研究室
Thesis Advisor夏仁波
ClassificationTP391.4
Keyword机场跑道识别 前下视红外图像 直线段提取 长度回溯 纹理相似性
Call NumberTP391.4/W94/2014
Pages59页
Degree Discipline机械电子工程
Degree Name硕士
2014-05-28
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract机场跑道识别具有重要的军事和民用价值,随着红外成像技术在精确制导领域中的广泛应用,红外图像机场跑道识别引起了国内外研究者的深入研究。目前提出的大多数机场识别算法都存在一定的局限性,只能处理一定成像距离下的图像,而且其特征提取算法在成像质量比较差的情况下不能稳定地获得对应的特征。为此,本文针对前下视红外图像,提出了一种适应不同成像距离并且特征提取稳定的机场跑道快速识别算法,主要内容包括:首先,研究了图像中直线段特征提取方法。分析了几种经典的直线段检测算法,如Hough变换、LSD、EDLines,并比较了它们在红外图像中的直线段检测效果,发现LSD算法在检测速度、直线段定位精度、虚警率控制、鲁棒性等方面取得了最好的综合效果。针对LSD算法检测的直线段存在断裂问题,本文提出了一种基于机场跑道边缘提取的直线段连接算法,它与LSD结合用于低信噪比、线性边缘模糊的前下视红外图像中直线段检测。实验结果表明,这两种方法结合能够适应不同分辨率的红外图像,并取得了满意的直线段提取效果。其次,根据机场的线状结构特征和纹理特征,提出了一种确定机场跑道完整候选区域(ROI)的算法。该算法先在已检测的直线段集合中提取满足跑道基本结构特征的平行直线段对,平行直线段对之间包含的区域即为跑道候选区域;然后基于跑道的纹理及结构特征,用本文提出的跑道长度回溯算法对所有ROI处理,提取出完整的跑道候选区域,很好地弥补了直线段提取算法在低信噪比图像中无法提取完整跑道边缘的不足。实验结果表明,该算法能够有效地提取出完整的跑道ROI,构造了一些目标假设,缩小了后续目标识别的范围,提高了识别效率。最后,在跑道ROI的基础上,定义了一些基于知识的机场跑道判别准则,依据这些准则从ROI中识别出真实的跑道目标。判别准则主要包括跑道的先验宽度约束、跑道及其邻域的灰度及平坦度对比、跑道区域自相似性以及最优候选跑道判决。为了适应不同成像距离下的目标识别,本文采取了一种分段识别的方法,即在近视图和远视图时,用不同的特征去定义判别准则来识别跑道,最终将二者融合到一起,这种识别方法具有更广泛的应用范围。综上所述,本文根据机场跑道的线状结构特征和区域灰度分布特性,提出了一种基于知识的跑道识别算法。该算法充分利用跑道线状特征的稳定性,利用直线段来定位出候选区域;然后利用跑道区域的灰度及结构特征,制定了跑道识别准则。实验结果表明,本文的识别算法适用于不同弹目距离下的机场跑道识别,识别率高,并且具有一定的实时性。
Other AbstractAirport runway recognition technology plays an important role both for military and civilian applications. With the extensive application of infrared imaging technique in the field of precision-guidance, recognition of airport runway based on infrared images instead of visible images has become a research hot topic around the world. But there exist some limitations in most of the proposed runway recognition algorithms, which can only handle the images under certain imaging distance or cannot get the corresponding image features in the case of poor image quality. In the paper, a new method of runway recognition is proposed for the forward-looking infrared (FLIR) images under different imaging distances and backgrounds. Firstly, several methods of line segments detection are analyzed, such as the method based on Hough Transform, the method of LSD and the EDLines method. From the results of line segments detection in the infrared images using these three methods, we can find that LSD gets the optimal comprehensive performance in the computation speed, positioning accuracy of line, false alarm rate and the detection robustness. But there is a drawback of LSD which generates fragmented line segments. Therefore, an improved line segment linking algorithm is introduced to connect the broken line segments aiming to extract the edges of the runway. The combination of LSD and the improved line segment linking algorithm can adapt to FLIR images with low SNR and with different resolutions. Afterwards, the method of extracting complete runway ROI is presented according to the linear structure feature and the texture feature of the runway. The runway has two long similar edges. Based on this, the parallel line segments are extracted as ROIs from the line segments set produced by the above line extraction method. Then a runway length backtracking method is proposed to retrieve the complete ROI based on the runway structure and texture features. This method compensates for the LSD in the case of processing very low SNR images or images with some interference that LSD can only extract local short lines of the runway. Experimental results show that our ROI extraction method can get the complete runway candidate region successfully and narrow the search scope for the subsequent processing.Finally, some decision criteria are formulated based on knowledge to get the true target in the ROIs. The priori knowledge of runway width, the contrast of gray and flatness between runway region and its neighborhood, the runway regional self-similarity are used to make the decision criteria. For adapting to images under different imaging distances, a piecewise constraint method is adopted to recognize the target. Different features are combined to formulate decision criteria when processing far view and close-up view images.By all accounts, the method proposed in the paper makes full use of the linear structure feature and the texture feature of the runway. Experimental results on the FLIR images with different imaging ranges demonstrate that this method is robust and has a good real-time performance.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/14794
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
武伟. 基于知识的前下视红外图像机场跑道识别算法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2014.
Files in This Item:
File Name/Size DocType Version Access License
基于知识的前下视红外图像机场跑道识别算法(2431KB) 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[武伟]'s Articles
Baidu academic
Similar articles in Baidu academic
[武伟]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[武伟]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.