【摘要】：近年來,國(guó)內(nèi)部分地區(qū)的霧霾污染日趨常態(tài)化,霧霾治理問題已成為社會(huì)各界密切關(guān)注的議題之一。霧霾污染及其所引起的能見度降低問題,不僅危害人類身心健康,而且給人類的室外活動(dòng)和交通出行造成了極大的不便。尤其是突發(fā)性的團(tuán)霧或者霧霾天氣極大地威脅著司機(jī)的行車安全。眾所周知,準(zhǔn)確的能見度檢測(cè)是解決該問題的必要環(huán)節(jié)之一,但現(xiàn)有設(shè)備和方法的性價(jià)比、準(zhǔn)確度和普及性有待提高。因此,著眼于人類的健康與出行,亟待一種實(shí)時(shí)且有效的霧霾能見度檢測(cè)方法。由于具有重要的理論和實(shí)用價(jià)值,霧霾能見度檢測(cè)方法已成為圖像處理和計(jì)算機(jī)視覺領(lǐng)域的研究熱點(diǎn),并引起了學(xué)者的廣泛關(guān)注。為了克服傳統(tǒng)能見度檢測(cè)方法的各種缺點(diǎn),研究人員結(jié)合攝像機(jī)標(biāo)定技術(shù)、圖像邊緣檢測(cè)和機(jī)器學(xué)習(xí)等方法,對(duì)基于數(shù)字圖像處理的能見度檢測(cè)方法進(jìn)行了研究。本文分析了能見度檢測(cè)相關(guān)原理,并利用監(jiān)控視頻圖像來檢測(cè)道路能見度,主要研究?jī)?nèi)容如下:1.本文針對(duì)暗通道先驗(yàn)理論估計(jì)透射率的不足之處,采用參數(shù)修正法優(yōu)化透射率�；谙袼攸c(diǎn)的亮通道理論來描述天空亮度函數(shù),減小獲取天空亮度時(shí)產(chǎn)生的誤差。同時(shí),采用導(dǎo)向?yàn)V波方法消除透射率圖中的塊狀效應(yīng)。最后,針對(duì)高速公路環(huán)境設(shè)計(jì)了一種快速的車道線檢測(cè)方法,通過獲取車道線端點(diǎn)信息來輔助估計(jì)能見度,大量的實(shí)驗(yàn)證明了該算法具有較好的檢測(cè)準(zhǔn)確率。2.本文在詳細(xì)論證了圖像熵用于霧霾能見度檢測(cè)的可行性后,針對(duì)高速公路的檢測(cè)環(huán)境,提出了一種基于最小圖像熵的能見度檢測(cè)算法。首先,本文對(duì)霧霾圖像提取道路區(qū)域并計(jì)算圖像的暗通道和透射率,利用圖像中的車道線先驗(yàn)信息計(jì)算圖像的場(chǎng)景深度信息。然后,根據(jù)大氣散射模型得到恢復(fù)圖像,計(jì)算出恢復(fù)圖像在道路區(qū)域的局部圖像熵。最后,通過搜索圖像熵的極小值對(duì)應(yīng)的消光系數(shù),即可得到該霧霾圖像的大氣能見度。實(shí)驗(yàn)結(jié)果證明,該算法符合人類視覺觀測(cè)效果,滿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.
【學(xué)位授予單位】：南京郵電大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41;P412.17;U491.53

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 宋洪軍;郜園園;陳陽舟;;基于攝像機(jī)動(dòng)態(tài)標(biāo)定的交通能見度估計(jì)[J];計(jì)算機(jī)學(xué)報(bào);2015年06期

2 李加元;胡慶武;艾明耀;嚴(yán)俊;;結(jié)合天空識(shí)別和暗通道原理的圖像去霧[J];中國(guó)圖象圖形學(xué)報(bào);2015年04期

3 郭庚山;葉青;胡鑫;;一種改進(jìn)的霧天圖像復(fù)原算法[J];計(jì)算機(jī)與現(xiàn)代化;2015年04期

4 陳萌;張紅英;吳亞東;刁揚(yáng)桀;;基于暗原色先驗(yàn)的夜視圖像增強(qiáng)算法[J];信息通信;2015年01期

5 崔寶俠;賈冬雪;段勇;;明亮區(qū)域的暗原色先驗(yàn)算法[J];沈陽工業(yè)大學(xué)學(xué)報(bào);2015年01期

6 宋得成;徐國(guó)慶;魯建勇;;暗原色先驗(yàn)圖像去霧改進(jìn)算法[J];武漢工程大學(xué)學(xué)報(bào);2014年12期

7 吳煒;李勃;楊嫻;端金鳴;陳啟美;;基于路面視亮度差平方最優(yōu)化的視頻能見度檢測(cè)算法[J];電子與信息學(xué)報(bào);2014年10期

8 郭尚書;齊文新;齊宇;;基于暗通道先驗(yàn)的視頻能見度測(cè)量方法[J];計(jì)算機(jī)與數(shù)字工程;2014年04期

9 和曉軍;喬寅;;基于暗原色的單一圖像去霧算法的研究[J];計(jì)算機(jī)應(yīng)用研究;2014年01期

10 石文軒;詹詩縈;李婕;;一種邊緣優(yōu)化的暗通道去霧算法[J];計(jì)算機(jī)應(yīng)用研究;2013年12期

相關(guān)碩士學(xué)位論文 前1條

1 苗苗;基于高清視頻的能見度檢測(cè)技術(shù)[D];北方工業(yè)大學(xué);2012年

，

本文編號(hào)：2296464