【作者】 黄冉希；

【导师】 桑农；

【作者基本信息】 华中科技大学 ， 模式识别与智能系统， 2006， 硕士

【摘要】 机场识别一直是人们长期关心的问题。它包括判断机场存在与否和机场定位两个方面。目前,国内外提出了大量的机场识别方法,但是在判断机场的存在性方面缺少有效的判断机制。其中,基于上下文的机场目标识别方法具有准确的机场定位性能,但是存在较高的虚警率。为此,本文提出了一种在保证较高探测概率的前提下,减少机场识别虚警率的方法。首先,本文介绍了基于上下文信息的机场目标识别方法,并且分别针对跑道检测,联络道检测和置信度3个方面,分析该方法产生虚警的原因。其次,进行基于跑道对比度特征的跑道识别。根据机场跑道目标具有亮度显著性这一特点,利用跑道直线与其它非跑道直线Radon变换值的对比性,提出了几个候选的跑道对比度特征。并且通过统计特性分析,选取了一种最能够突出对比性的候选量作为跑道对比度特征。然后,对每幅待判断图像计算其对比度特征,提出了两种判别准则:类间差异和类内方差准则。并通过实验比较了这两种判别准则的优缺点。在这一步中,如若图像被判为不存在机场跑道,则该图像中不存在机场。最后,在初步判断存在机场的图像中,通过联络道识别,进一步减少虚警。联络道识别主要是利用联络道灰度的均匀性、亮度显著性和联络道延长线信息,这3个特性来判断联络道的真伪。如果存在至少一条联络道,则说明图像中存在机场。实验结果表明,大部分虚警都能在基于跑道对比度特征的识别时被排除,联络道识别也具有一定的减少虚警的能力。更多还原

【Abstract】 Airport recognition has always received much attention. It contains two aspects: identification and localization of airport. Nowadays, many algorithms for airport recognition have been presented, but lack of efficient strategies to identify airport. Among them, an algorithm based on contextual information has been approved to exactly detect the airport localization, but has a high false alarm rate. So, a new method to reduce false alarm in the airport recognition under the precondition of high detection rate is presented.Firstly, this thesis introduces the algorithm to recognize airport based on contextual information, and then analyzes the origins of high false alarm rate from 3 aspects: runway detection, taxiway detection and the criteria of airport identification.Secondly, runway recognition using a runway contrast metric is conducted. Since runways are always significantly bright, we use the contrast between runway lines’and other lines’Radon transformed results to characterize the runway contrast metric. We present several candidate metrics and choose one that can best express the contrast.Thirdly, two discriminances to classify the runway contrast metric are presented. One is based on the difference of the two image classes (airport image and non-airport image), the other is based on the diversity of non-airport image. We compare the two discriminances through experiments.Finally, taxiway recognition is used to further reduce false alarm after runway recognition. Taxiway recognition is mainly according to 3 features: the uniformity of taxiway’s grayscale, the salience of taxiway’s brightness and the information of taxiway’s extended lines. Recognition of at least one taxiway can prove the existence of airport.The experiments demonstrate most false alarms can be excluded in the runway recogniton and the rest can be eliminated in the taxiway recognition.更多还原