本文選題：智能交通 + 交通事件檢測(cè)��； 參考：《華南理工大學(xué)》2015年碩士論文

【摘要】：視頻交通事件檢測(cè)系統(tǒng)是利用圖像處理技術(shù),結(jié)合人工智能和機(jī)器學(xué)習(xí)等學(xué)科知識(shí)對(duì)高速公路監(jiān)控視頻進(jìn)行分析,自動(dòng)檢測(cè)出高速公路上出現(xiàn)的異常事件的系統(tǒng)。它能夠快速、準(zhǔn)確的檢測(cè)出高速公路運(yùn)營(yíng)過(guò)程中出現(xiàn)的異常事件,同時(shí)向高速公路管理人員發(fā)出相應(yīng)信息,將異常事件對(duì)高速公路運(yùn)營(yíng)的影響降到最低,確保高速公路暢通運(yùn)行。它是高速公路應(yīng)急管理平臺(tái)及智能交通系統(tǒng)中不可或缺的前端信息采集子系統(tǒng)。隨著我國(guó)高速公路擁堵問(wèn)題日益嚴(yán)重,視頻交通事件檢測(cè)系統(tǒng)正逐步得到廣泛的應(yīng)用,因此關(guān)于此技術(shù)的研究就顯得非常有意義。結(jié)合高速公路監(jiān)控系統(tǒng)現(xiàn)狀,本文對(duì)目前監(jiān)控系統(tǒng)中存在的問(wèn)題進(jìn)行分析,擬解決幾個(gè)當(dāng)前視頻檢測(cè)系統(tǒng)中普遍存在的問(wèn)題,著重提高視頻檢測(cè)系統(tǒng)的檢測(cè)率和環(huán)境穩(wěn)定性。本文首先解決的問(wèn)題就是提高監(jiān)控系統(tǒng)的環(huán)境適應(yīng)能力。由于高速公路監(jiān)控大多屬于露天環(huán)境,監(jiān)控環(huán)境復(fù)雜多變,對(duì)監(jiān)控系統(tǒng)的環(huán)境適應(yīng)能力有較高的要求,通過(guò)對(duì)多條高速公路現(xiàn)場(chǎng)環(huán)境分析,采集數(shù)據(jù),本文提出了一種新的分類算法對(duì)日夜切換、雨天路面積水、夜間雨天路燈反光等惡劣環(huán)境進(jìn)行識(shí)別。其次,利用背景差分技術(shù)對(duì)高速公路中存在的交通目標(biāo)進(jìn)行目標(biāo)檢測(cè)。在背景差分二值化目標(biāo)檢測(cè)技術(shù)中,提出了一種新的自適應(yīng)背景差分閾值確定算法,以適應(yīng)高速公路復(fù)雜多變的監(jiān)控環(huán)境。同時(shí)針對(duì)惡劣監(jiān)控環(huán)境采用有針對(duì)性的識(shí)別算法以保證系統(tǒng)檢測(cè)的準(zhǔn)確性和穩(wěn)定性。最后,利用基于目標(biāo)跟蹤檢測(cè)的事件檢測(cè)系統(tǒng)對(duì)獲取的目標(biāo)進(jìn)行跟蹤檢測(cè)分析。構(gòu)建了適應(yīng)于高速公路監(jiān)控環(huán)境運(yùn)動(dòng)目標(biāo)跟蹤的Kalman預(yù)估器,分析各類異常交通事件中運(yùn)動(dòng)目標(biāo)特征或者監(jiān)控畫面的特點(diǎn),通過(guò)機(jī)器學(xué)習(xí)算法對(duì)各類常見的交通事件的特征參數(shù)進(jìn)行定義及閾值處理,實(shí)現(xiàn)了各類異常交通事件的準(zhǔn)確檢測(cè)。通過(guò)在國(guó)內(nèi)某幾條高速公路進(jìn)行的模擬測(cè)試驗(yàn)證了本文提出的交通事件檢測(cè)算法具有很好的性能,在多種復(fù)雜環(huán)境下均能準(zhǔn)確檢測(cè)出各類常見異常事件。
[Abstract]:Video traffic event detection system is a system that uses image processing technology, combining with artificial intelligence and machine learning to analyze the video of highway surveillance, and automatically detects the abnormal events on the highway. It can detect the abnormal events in the expressway operation process quickly and accurately, at the same time, it can send the corresponding information to the expressway management personnel to minimize the impact of the abnormal events on the expressway operation. Make sure the freeway runs smoothly. It is an indispensable front-end information collection subsystem in expressway emergency management platform and intelligent transportation system. With the increasingly serious problem of highway congestion in China, video traffic incident detection system is gradually being widely used, so the research on this technology is very meaningful. Combined with the current situation of highway monitoring system, this paper analyzes the existing problems in the current monitoring system, and proposes to solve several common problems in the current video detection system, focusing on improving the detection rate and environmental stability of the video detection system. The first problem to be solved in this paper is to improve the environmental adaptability of the monitoring system. Because the expressway monitoring mostly belongs to the open-air environment, the monitoring environment is complex and changeable, and has higher requirements for the environmental adaptability of the monitoring system. In this paper, a new classification algorithm is proposed to identify the bad environment, such as day and night switching, rainy day road surface water accumulation, night rainy day street lamp reflection and so on. Secondly, the background differential technique is used to detect the traffic targets in expressway. In the background differential binary target detection technology, a new adaptive background differential threshold determination algorithm is proposed to adapt to the complicated and changeable monitoring environment of freeway. In order to ensure the accuracy and stability of the detection system, a targeted recognition algorithm is adopted for the adverse monitoring environment at the same time. Finally, the event detection system based on target tracking detection is used to track and detect the acquired target. A Kalman predictor suitable for tracking moving targets in highway monitoring environment is constructed to analyze the characteristics of moving targets or monitoring pictures in various kinds of abnormal traffic events. By means of machine learning algorithm, the characteristic parameters of various common traffic events are defined and the threshold value is processed to realize the accurate detection of all kinds of abnormal traffic events. The traffic event detection algorithm proposed in this paper is proved to have good performance by simulating tests on some domestic freeways. It can accurately detect all kinds of common abnormal events in various complex environments.
【學(xué)位授予單位】：華南理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：U495;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 王海麗;張良;;基于相關(guān)區(qū)域約束的SURF特征點(diǎn)匹配[J];通信技術(shù);2012年02期

2 于洋,劉二莉,周鐵濤,劉培英;邊界追蹤及Freeman碼在定量金相中的應(yīng)用[J];北京航空航天大學(xué)學(xué)報(bào);2004年08期

3 樓建光,柳崎峰,譚鐵牛,胡衛(wèi)明;基于三維模型的交通場(chǎng)景視覺監(jiān)控(英文)[J];自動(dòng)化學(xué)報(bào);2003年03期

相關(guān)博士學(xué)位論文 前1條

1 胡宏宇;基于視頻處理的交通事件識(shí)別方法研究[D];吉林大學(xué);2010年

，

本文編號(hào)：2069535