基于道路监控视频的雾霾能见度检测方法研究

发布时间：2018-10-26 17:32

【摘要】：近年来,国内部分地区的雾霾污染日趋常态化,雾霾治理问题已成为社会各界密切关注的议题之一。雾霾污染及其所引起的能见度降低问题,不仅危害人类身心健康,而且给人类的室外活动和交通出行造成了极大的不便。尤其是突发性的团雾或者雾霾天气极大地威胁着司机的行车安全。众所周知,准确的能见度检测是解决该问题的必要环节之一,但现有设备和方法的性价比、准确度和普及性有待提高。因此,着眼于人类的健康与出行,亟待一种实时且有效的雾霾能见度检测方法。由于具有重要的理论和实用价值,雾霾能见度检测方法已成为图像处理和计算机视觉领域的研究热点,并引起了学者的广泛关注。为了克服传统能见度检测方法的各种缺点,研究人员结合摄像机标定技术、图像边缘检测和机器学习等方法,对基于数字图像处理的能见度检测方法进行了研究。本文分析了能见度检测相关原理,并利用监控视频图像来检测道路能见度,主要研究内容如下:1.本文针对暗通道先验理论估计透射率的不足之处,采用参数修正法优化透射率。基于像素点的亮通道理论来描述天空亮度函数,减小获取天空亮度时产生的误差。同时,采用导向滤波方法消除透射率图中的块状效应。最后,针对高速公路环境设计了一种快速的车道线检测方法,通过获取车道线端点信息来辅助估计能见度,大量的实验证明了该算法具有较好的检测准确率。2.本文在详细论证了图像熵用于雾霾能见度检测的可行性后,针对高速公路的检测环境,提出了一种基于最小图像熵的能见度检测算法。首先,本文对雾霾图像提取道路区域并计算图像的暗通道和透射率,利用图像中的车道线先验信息计算图像的场景深度信息。然后,根据大气散射模型得到恢复图像,计算出恢复图像在道路区域的局部图像熵。最后,通过搜索图像熵的极小值对应的消光系数,即可得到该雾霾图像的大气能见度。实验结果证明,该算法符合人类视觉观测效果,满足高速公路安全要求。
[Abstract]:In recent years, haze pollution in some parts of China has become more and more regular, and haze treatment has become one of the issues concerned by all walks of life. The pollution of haze and the reduced visibility caused by it not only endanger human physical and mental health, but also cause great inconvenience to human outdoor activities and transportation. In particular, sudden fog or haze weather is a major threat to driver safety. It is well known that accurate visibility detection is one of the necessary links to solve this problem, but the cost performance, accuracy and popularization of existing equipment and methods need to be improved. Therefore, in view of human health and travel, it is urgent to develop a real-time and effective haze visibility detection method. Because of its important theoretical and practical value, haze visibility detection method has become a research hotspot in the field of image processing and computer vision, and has attracted extensive attention of scholars. In order to overcome the shortcomings of the traditional visibility detection methods, the visibility detection method based on digital image processing is studied by the researchers in combination with camera calibration, image edge detection and machine learning. This paper analyzes the related principles of visibility detection, and uses video surveillance images to detect road visibility. The main research contents are as follows: 1. Aiming at the shortcomings of dark channel priori theory in estimating transmittance, a parameter correction method is used to optimize the transmittance. Based on the bright channel theory of pixels, the sky brightness function is described to reduce the error when the sky brightness is obtained. At the same time, guided filter is used to eliminate the block effect in the transmittance map. Finally, a fast lane detection method is designed for freeway environment, which can obtain lane endpoint information to help estimate visibility. A large number of experiments show that the algorithm has a good detection accuracy. 2. In this paper, the feasibility of applying image entropy to haze visibility detection is demonstrated in detail, and a visibility detection algorithm based on minimum image entropy is proposed for highway detection environment. Firstly, the road area is extracted from the haze image and the dark channel and transmittance of the image are calculated, and the scene depth information is calculated by using the prior information of the lane line in the image. Then, according to the atmospheric scattering model, the restoration image is obtained, and the local image entropy of the restored image in the road region is calculated. Finally, the atmospheric visibility of the haze image can be obtained by searching the extinction coefficient corresponding to the minimum value of the image entropy. The experimental results show that the algorithm accords with the human visual observation effect and meets the highway safety requirements.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;P412.17;U491.53

【参考文献】