【摘要】：由于國(guó)內(nèi)交通快速發(fā)展,引起的負(fù)面影響就是交通事故急劇增加,其中有許多由于車道偏離引起的交通事故,因此實(shí)時(shí)性高、可靠性強(qiáng)的車道線檢測(cè)與跟蹤成為了車輛導(dǎo)航性能要求的主要內(nèi)容。近年來(lái),由于許多研究人員的努力,在這一領(lǐng)域已經(jīng)取得了一些進(jìn)展,例如,應(yīng)用在高速公路場(chǎng)景中的道路識(shí)別已經(jīng)非常成熟。本文對(duì)車道線的檢測(cè)與跟蹤進(jìn)行了研究,其中車道線檢測(cè)被廣泛應(yīng)用于自動(dòng)駕駛和防撞報(bào)警系統(tǒng)中。車道線檢測(cè)系統(tǒng)即在道路圖像中,通過(guò)預(yù)處理算法排除干擾,以及初步的對(duì)圖像進(jìn)行整理,提取出有效的車道線信息,且將其識(shí)別。本文車道線檢測(cè)主要包括預(yù)處理、車道線檢測(cè)算法、車道線跟蹤算法以及CSK算法改進(jìn)四部分。(1)車道線檢測(cè)算法。通過(guò)預(yù)處理算法,經(jīng)過(guò)逆透視變換,高斯濾波以及分位數(shù)方法,對(duì)于不同光照亮度的圖像實(shí)施車道線檢測(cè)做準(zhǔn)備。然后對(duì)于不同光照亮度的圖像做灰度化處理,車道線識(shí)別主要運(yùn)用了改善的快速隨機(jī)抽樣的線性擬合一致性。(2)車道線跟蹤算法。本文研究了卡爾曼濾波算法和CSK(Exploiting the Circulant Structure of Tracking-by-Detection with Kernels)跟蹤算法,同時(shí)發(fā)現(xiàn)CSK算法在目標(biāo)被遮擋時(shí),無(wú)法實(shí)現(xiàn)跟蹤,為此對(duì)CSK算法不防遮擋進(jìn)行了改進(jìn)研究。(3)算法測(cè)試。根據(jù)本文的算法,車道檢測(cè)和跟蹤測(cè)試的真實(shí)場(chǎng)景。檢測(cè)結(jié)果表明,所提出的檢測(cè)算法能夠準(zhǔn)確、快速的實(shí)現(xiàn)車道線檢測(cè);跟蹤結(jié)果表明,卡爾曼濾波以及CSK跟蹤算法對(duì)比分析實(shí)驗(yàn)數(shù)據(jù),CSK比Kalman跟蹤的效率高,速度快,而且改進(jìn)后的CSK算法能夠成功實(shí)現(xiàn)遮擋時(shí)目標(biāo)的跟蹤。
[Abstract]:Due to the rapid development of domestic traffic, the negative impact is the sharp increase of traffic accidents, many of which are caused by lane deviation, so the real-time performance is high. High reliability lane detection and tracking has become the main content of vehicle navigation performance requirements. In recent years, due to the efforts of many researchers, some progress has been made in this field, for example, the road recognition used in freeway scene is very mature. In this paper, the detection and tracking of lane lines are studied, in which lane detection is widely used in automatic driving and collision alarm systems. Lane detection system in the road image, through the pre-processing algorithm to eliminate interference, and preliminary collation of the image, extract effective lane information, and identify it. This paper mainly includes four parts: pretreatment, lane detection algorithm, lane tracking algorithm and improved CSK algorithm. (1) Lane line detection algorithm. Through pre-processing algorithm, inverse perspective transformation, Gao Si filter and quantile method, the lane detection of images with different illumination brightness is prepared. Then the grayscale image with different illumination brightness is processed, the lane line recognition mainly uses the improved linear fitting consistency of the fast random sampling. (2) the lane line tracking algorithm. In this paper, the Kalman filter algorithm and CSK (Exploiting the Circulant Structure of Tracking-by-Detection with Kernels) tracking algorithm are studied, and it is found that the CSK algorithm can not achieve tracking when the target is occluded. Therefore, the CSK algorithm is improved. (3) algorithm test. According to this algorithm, lane detection and tracking test of the real scene. The results show that the proposed algorithm can detect the lane accurately and quickly, and the tracking results show that the Kalman filter and the CSK tracking algorithm are more efficient and faster than the Kalman tracking algorithm in comparing and analyzing the experimental data. Moreover, the improved CSK algorithm can successfully achieve occlusion target tracking.
【學(xué)位授予單位】：長(zhǎng)安大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：U463.6;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前6條

1 楊益;何穎;;基于RGB空間的車道線檢測(cè)與辨識(shí)方法[J];計(jì)算機(jī)與現(xiàn)代化;2014年02期

2 肖延勝;;為極速智能車保駕護(hù)航——記清華大學(xué)計(jì)算機(jī)系THMR課題組之智能汽車研究[J];中國(guó)發(fā)明與專利;2011年12期

3 張曦;黃亮;徐洋;郭冬霄;;基于MATLAB中calibration toolbox的相機(jī)標(biāo)定應(yīng)用研究[J];微型機(jī)與應(yīng)用;2011年14期

4 高德芝;鄭榜貴;段建民;;基于逆透視變換的智能車輛定位技術(shù)[J];計(jì)算機(jī)測(cè)量與控制;2009年09期

5 雷濤;樊養(yǎng)余;黃連冰;;Fast lane recognition based on morphological multi-structure element model[J];Optoelectronics Letters;2009年04期

6 孫振平,安向京,賀漢根;CITAVT-IV——視覺(jué)導(dǎo)航的自主車[J];機(jī)器人;2002年02期

相關(guān)會(huì)議論文 前1條

1 查宇飛;張育;畢篤彥;;一種基于粒子濾波的自適應(yīng)運(yùn)動(dòng)目標(biāo)跟蹤方法[A];第十二屆全國(guó)圖象圖形學(xué)學(xué)術(shù)會(huì)議論文集[C];2005年

相關(guān)碩士學(xué)位論文 前5條

1 謝茜;基于視覺(jué)的車道線檢測(cè)與識(shí)別[D];武漢理工大學(xué);2013年

2 孫占瑞;基于模糊聚類與模糊模式識(shí)別的車道偏離預(yù)警方法[D];燕山大學(xué);2012年

3 梁雄;自主導(dǎo)航車局部地圖創(chuàng)建研究[D];中南大學(xué);2011年

4 馬超;基于單目視覺(jué)的車道偏離預(yù)警系統(tǒng)設(shè)計(jì)[D];電子科技大學(xué);2011年

5 孟海崗;基于平面約束的CCD相機(jī)標(biāo)定方法改進(jìn)[D];吉林大學(xué);2009年

，

本文編號(hào)：2201541