本文選題：交通視頻　切入點(diǎn)：車輛檢測　出處：《湖南大學(xué)》2014年碩士論文

【摘要】：實(shí)時采集道路網(wǎng)絡(luò)中的道路交通流量、車速、車流密度、車頭時距、車輛行程時間、以及占用率等宏觀和微觀的道路交通信息，可以幫助掌握道路交通流的規(guī)律，指導(dǎo)人們進(jìn)行道路交通管理與控制。道路交通信息是道路交通設(shè)施改善、道路交通規(guī)劃等必要的基礎(chǔ)數(shù)據(jù)。因此，建立道路交通信息采集方法是一個很值得研究的課題。本文的主要研究內(nèi)容如下： （1）采用統(tǒng)計方法更新視頻圖像背景，同時為了適應(yīng)道路交通的復(fù)雜性，對不同道路環(huán)境下背景更新過程中相關(guān)參數(shù)進(jìn)行了討論和分析，并給出了參考值。在獲取道路背景后，利用鄰域、多閾值的方法對傳統(tǒng)背景差分法來進(jìn)行改進(jìn)，克服其在運(yùn)動目標(biāo)檢測過程中的不足。然后通過融合圖像陰影的多種特征，，獲得包含視頻圖像陰影和亮度較低車輛的區(qū)域，再構(gòu)建一組向量消除視頻圖像陰影的影響。通過詳細(xì)的實(shí)驗進(jìn)一步驗證采用改進(jìn)的背景差分法和本文提出的陰影消除方法對車輛檢測結(jié)果的可靠性。 （2）為消除視頻圖像中車輛之間相互遮擋的影響，提高道路交通信息檢測的準(zhǔn)確性，提出了一種基于數(shù)學(xué)形態(tài)學(xué)的方法對視頻圖像中車輛之間相互遮擋的圖像進(jìn)行分割。計算視頻圖像中的車輛實(shí)體面積與最小外接多邊形的面積差，以及視頻圖像中單個連通區(qū)域內(nèi)車輛面積來智能識別遮擋是否存在。存在遮擋時，通過鄰域特征搜索到凹包的頂點(diǎn)，然后利用凹包的頂點(diǎn)，作為腐蝕起點(diǎn)的判斷依據(jù)，再腐蝕掉在距離變換過程中受凹包頂點(diǎn)影響的像素點(diǎn)。在完成對目標(biāo)車輛遮擋識別和相關(guān)處理后，本文融合基于特征匹配的追蹤方法和基于灰色系統(tǒng)的預(yù)測方法對運(yùn)動目標(biāo)追蹤，先利用特征匹配追蹤方法對運(yùn)動車輛進(jìn)行初步追蹤，并利用灰色預(yù)測進(jìn)行輔助判斷特征匹配結(jié)果的可靠性。 （3）采用計算機(jī)編程將相關(guān)理論和技術(shù)轉(zhuǎn)化為實(shí)踐，本文設(shè)計并實(shí)現(xiàn)了道路交通信息采集的軟件。選取了代表近距離拍攝和遠(yuǎn)距離拍攝的兩個交通視頻實(shí)例對該技術(shù)進(jìn)行檢驗和分析，獲取了交通量、車速、車頭時距、交通密度，車輛達(dá)到時間等交通信息。
[Abstract]:The real-time collection of macro and micro road traffic information, such as road traffic flow, speed, vehicle density, headway time, vehicle travel time, and occupancy rate, can help to master the laws of road traffic flow. Guiding people to carry out road traffic management and control. Road traffic information is the necessary basic data such as road traffic facilities improvement, road traffic planning and so on. The establishment of road traffic information collection method is a topic worthy of study. The main contents of this paper are as follows:. In order to adapt to the complexity of road traffic, the relevant parameters in the background updating process under different road environments are discussed and analyzed, and the reference values are given. The traditional background difference method is improved by using neighborhood and multi-threshold method to overcome the shortcomings in moving target detection. To obtain an area containing shadow and low brightness vehicles in the video image, Then a set of vectors is constructed to eliminate the shadow effect of video images. The reliability of the improved background difference method and the shadow cancellation method proposed in this paper is further verified by the detailed experiments. In order to eliminate the influence of mutual occlusion between vehicles in video images and improve the accuracy of road traffic information detection, In this paper, a mathematical morphology based method is proposed to segment the images of mutual occlusion between vehicles in video images, and to calculate the difference between the area of vehicles in video images and the minimum external polygon. When there is occlusion, the vertex of concave packet is searched by neighborhood feature, and then the vertex of concave packet is used as the basis for judging the corrosion starting point. Then erode the pixels affected by the concave vertex during the distance transformation. After the target vehicle occlusion recognition and correlation processing are completed, In this paper, the tracking method based on feature matching and the prediction method based on gray system are combined to track the moving target. Firstly, the tracking method of feature matching is used to track the moving vehicle. The reliability of feature matching results is evaluated by grey prediction. Using computer programming to translate relevant theories and technologies into practice, In this paper, the software of road traffic information collection is designed and implemented. Two traffic video examples, which represent close shooting and long distance shooting, are selected to test and analyze the technology, and the traffic volume, speed, headway time distance and traffic density are obtained. Vehicle arrival time and other traffic information.
【學(xué)位授予單位】：湖南大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：U495

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 吳成東;郭利鋒;張云洲;劉o

本文編號：1685574