基于视频图像的能见度检测算法的研究

发布时间：2018-07-27 12:19

【摘要】：能见度是一个重要的气象观测指标,低能见度对气象、交通、军事和农业等领域影响很大。其中对交通领域的影响最大,高速公路经常发生由低能见度天气引发的交通事故,相当一部分都是多车连环相撞的重大交通事故,这对人们的生命和财产安全造成了很大的威胁。为了解决这个问题,需要在道路上密集布设能见度检测设备,及时对低能见度情况进行预警,保障交通运行的安全,传统能见度仪操作复杂、价格昂贵,不能满足要求。视频图像能见度检测具有成本低、操作简单和可靠性高等优点受到了很多研究者的关注。论文在现有研究成果的基础上,对一些视频图像能见度算法做了进一步的研究和改进,论文的主要研究工作如下:研究了双亮度差法,用这种算法实现了能见度的检测,并做了误差分析,当目标物黑体背后有遮挡物且背景天空灰度分布不均匀时,传统双亮度差法会出现较大检测误差,论文针对这个问题做了算法改进,用区域生长法分割出图像的天空区域,并对天空区域像素灰度求均值,实验证明该改进算法可以得到更准确的能见度值。研究了暗通道先验法,利用暗通道理论求出透射率,并用摄像机标定技术进行三维重建,得到距离信息,计算出能见度。针对用传统暗通道先验理论估算出的透射率天空区域的值偏小,提出一种用k-means聚类算法自适应分割阈值的方法来准确分割出天空区域,并修正天空区域透射率。对估算透射率过程中需要计算的大气光亮度值,当图像中有纯白色物体干扰时,用传统方法计算会出现估计值比实际值大,论文提出用阈值法估算大气光亮度。实验证明这些改进可以使能见度检测结果更加准确。对双亮度差法、改进的双亮度差法、暗通道先验法和改进暗通道先验法做能见度检测实验,用几种不同天气的视频图像数据,验证了改进的双亮度差法和改进的暗通道先验法都比其对应的传统算法检测结果更准确、可靠,而这些算法综合比较后得出,能见度检测准确度最高的是改进的暗通道先验法,该改进算法能适应各种不同天气。
[Abstract]:Visibility is an important meteorological observation index. Low visibility has great influence on meteorological, traffic, military and agricultural fields. Among them, it has the greatest impact on the traffic field. Traffic accidents caused by low-visibility weather often occur on expressways. Quite a number of major traffic accidents are caused by multi-vehicle pile-ups. This poses a great threat to the safety of people's lives and property. In order to solve this problem, it is necessary to place the visibility detection equipment on the road densely, to warn the low visibility situation in time, to ensure the safety of traffic operation. The traditional visibility meter is complex, expensive and unable to meet the requirements. Visibility detection of video images has attracted many researchers' attention because of its low cost, simple operation and high reliability. Based on the existing research results, this paper makes further research and improvement on some visibility algorithms of video images. The main research work of this paper is as follows: the double brightness difference method is studied, and the visibility detection is realized by this algorithm. The error analysis is done. When there is an occlusion behind the blackbody and the background sky gray distribution is not uniform, the traditional double brightness difference method will have a large detection error. In this paper, the algorithm is improved to solve this problem. The region growing method is used to segment the sky region of the image, and the average value of the pixel gray level in the sky region is obtained. The experimental results show that the improved algorithm can get more accurate visibility value. The dark channel priori method is studied and the transmittance is calculated by using the dark channel theory. The range information is obtained and the visibility is calculated by the camera calibration technique. In view of the small value of the transmittance sky region estimated by the traditional dark channel priori theory, an adaptive threshold segmentation method based on k-means clustering algorithm is proposed to segment the sky region accurately and to correct the transmittance of the sky region. For the atmospheric luminance value which needs to be calculated in the process of estimating transmittance, when there is interference of pure white object in the image, the estimated value by traditional method will be larger than the actual value. In this paper, the threshold method is proposed to estimate atmospheric luminance. Experiments show that these improvements can make the visibility detection results more accurate. For the double brightness difference method, the improved double brightness difference method, the dark channel priori method and the improved dark channel prior method, the visibility detection experiments are done, and the video image data of several different weather conditions are used. The improved double brightness difference method and the improved dark channel priori method are verified to be more accurate and reliable than the corresponding traditional algorithms. The improved algorithm can adapt to different weather conditions.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41

【相似文献】