本文選題：智能交通 + 交通事故檢測(cè)�。� 參考：《電子科技大學(xué)》2017年碩士論文

【摘要】：交通事故檢測(cè)是智能交通系統(tǒng)中最重要的部分之一,實(shí)時(shí)且魯棒的交通事故檢測(cè)方法可以在減少人員傷亡和減少財(cái)產(chǎn)損失上做出巨大貢獻(xiàn)。隨著智能交通系統(tǒng)的快速發(fā)展,基于計(jì)算機(jī)視覺技術(shù)和圖像處理技術(shù)的交通事故檢測(cè)系統(tǒng)的研究吸引了廣泛的注意力,許多研究者在這個(gè)領(lǐng)域中也取得了重大進(jìn)展。然而,由于交通環(huán)境的復(fù)雜性,目前提出的一些方法在實(shí)際應(yīng)用中仍然存在一定的限制。為了保證對(duì)交通事故快速和精確地檢測(cè),并且達(dá)到實(shí)際運(yùn)用的目的,一些挑戰(zhàn)性的問題就需要被解決。這樣的一種交通事故檢測(cè)算法必須達(dá)到三個(gè)要求。第一,能夠處理復(fù)雜的交通環(huán)境,包括較差的天氣、不同的光照條件、復(fù)雜的道路狀況和不同的交通參與者;第二,能夠處理不同的交通流狀態(tài),也就是不同程度的交通擁堵;第三,能夠?qū)崟r(shí)性運(yùn)行。在本文中,我們研究在交通視頻流中實(shí)時(shí)和魯棒的交通事故檢測(cè)算法。傳統(tǒng)的方法要么就是運(yùn)行速度不夠快,要么就是在復(fù)雜的交通環(huán)境下魯棒性不夠好。我們提出了一種新穎的基于監(jiān)控視頻的交通事故檢測(cè)方法,該方法的主要觀察點(diǎn)在于交通事故的發(fā)生不僅會(huì)引起局部目標(biāo)的運(yùn)動(dòng)方向混亂,也會(huì)造成全局交通流的混亂。在我們的方法中,一共有三步。第一,對(duì)每一幀視頻構(gòu)建對(duì)應(yīng)的光流場(chǎng);第二,基于光流場(chǎng)構(gòu)建全局交通流描述子高斯模型和局部運(yùn)動(dòng)方向圖高斯模型,來對(duì)交通事故的全局特征和局部特征進(jìn)行檢測(cè),從而對(duì)視頻幀中發(fā)生的交通事故進(jìn)行檢測(cè)與定位;第三,構(gòu)建了一個(gè)檢驗(yàn)?zāi)K來進(jìn)一步驗(yàn)證交通事故的發(fā)生,并排除事故誤報(bào)的情況。我們的方法具有實(shí)時(shí)性運(yùn)行、高精度、漏報(bào)少、誤報(bào)率低以及對(duì)不同交通環(huán)境和光照條件具有魯棒性等優(yōu)點(diǎn),后期通過廣泛的量化評(píng)估實(shí)驗(yàn)驗(yàn)證了以上優(yōu)點(diǎn),并且展示了在這個(gè)領(lǐng)域中的巨大進(jìn)步。
[Abstract]:Traffic accident detection is one of the most important parts in intelligent transportation system. The real-time and robust traffic accident detection method can make a great contribution to reducing casualties and property losses. With the rapid development of intelligent transportation system, the research of traffic accident detection system based on computer vision and image processing technology has attracted extensive attention, and many researchers have also made great progress in this field. However, due to the complexity of traffic environment, there are still some limitations in the practical application of some proposed methods. In order to ensure the rapid and accurate detection of traffic accidents and to achieve the purpose of practical application, some challenging problems need to be solved. Such a traffic accident detection algorithm must meet three requirements. Firstly, it can deal with complicated traffic environment, including bad weather, different light conditions, complex road conditions and different traffic participants; second, it can deal with different traffic flow states, that is, traffic congestion of different degrees. Third, it can run in real time. In this paper, we study real-time and robust traffic accident detection algorithms in traffic video streams. The traditional approach is either not fast enough or robust enough in complex traffic environments. We propose a novel method for traffic accident detection based on surveillance video. The main observation point of this method is that the occurrence of traffic accident will not only cause confusion of the moving direction of local target, but also cause chaos of global traffic flow. There are three steps in our method. Firstly, the corresponding optical flow field is constructed for each frame of video; secondly, the global traffic flow descriptor Gao Si model and the local motion pattern Gao Si model are constructed based on the optical flow field to detect the global and local characteristics of traffic accidents. In order to detect and locate the traffic accident in the video frame, a verification module is constructed to further verify the occurrence of the traffic accident and eliminate the false alarm of the accident. Our method has the advantages of real-time operation, high precision, low false alarm rate, low false alarm rate and robustness to different traffic environment and illumination conditions. And showed great progress in this field.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：U495;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前9條

1 苑瑋琦;謝昌隆;;基于機(jī)器視覺的交通事件檢測(cè)方法的研究[J];計(jì)算機(jī)仿真;2013年10期

2 趙有婷;李熙瑩;羅東華;;基于視頻車輛軌跡模型的交通事件自動(dòng)檢測(cè)方法研究[J];中山大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年04期

3 田秋芳;陳陽舟;張利國(guó);;城市快速路交通事件檢測(cè)的自適應(yīng)算法研究[J];交通運(yùn)輸工程與信息學(xué)報(bào);2010年04期

4 嚴(yán)兆斌;方敏;;基于序列模式挖掘的公路隧道交通事件檢測(cè)[J];計(jì)算機(jī)應(yīng)用與軟件;2010年01期

5 陳小鴻;李兵;;基于小波變換的城市道路交通事件檢測(cè)[J];交通信息與安全;2009年06期

6 汪勤;黃山;張洪斌;楊權(quán);張建軍;;基于視頻圖像處理的交通事件檢測(cè)系統(tǒng)[J];計(jì)算機(jī)應(yīng)用;2008年07期

7 彭娟春;顧立忠;蘇劍波;;基于Camshift和Kalman濾波的仿人機(jī)器人手勢(shì)跟蹤[J];上海交通大學(xué)學(xué)報(bào);2006年07期

8 巨永鋒,潘若禹,李磊;數(shù)據(jù)融合技術(shù)在高速公路異常事件檢測(cè)中的應(yīng)用[J];現(xiàn)代電子技術(shù);2005年17期

9 呂琪,王慧;基于動(dòng)態(tài)神經(jīng)網(wǎng)絡(luò)模型的交通事件檢測(cè)算法[J];公路交通科技;2003年06期

，

本文編號(hào)：1960494