本文選題：隧道　切入點(diǎn)：行人檢測(cè)　出處：《昆明理工大學(xué)》2017年碩士論文　論文類(lèi)型：學(xué)位論文

【摘要】：隧道區(qū)域是高速公路管理的重點(diǎn)區(qū)域,行人和非機(jī)動(dòng)車(chē)輛違規(guī)進(jìn)入高速公路隧道內(nèi)會(huì)嚴(yán)重影響高速公路的正常運(yùn)行,造成巨大的安全隱患。因此,針對(duì)隧道環(huán)境下視頻監(jiān)控中的行人檢測(cè)技術(shù)是高速公路正常運(yùn)營(yíng)的重要保障。隧道環(huán)境內(nèi),環(huán)境光照條件差,在圖像中產(chǎn)生大量噪聲,行人在隧道內(nèi)目標(biāo)小,像素低,給隧道環(huán)境下行人檢測(cè)帶來(lái)很大挑戰(zhàn)。本文主要研究了視頻檢測(cè)中的前景目標(biāo)與背景目標(biāo)的分割方法,使用了基于數(shù)學(xué)特征提取方法與卷積神經(jīng)網(wǎng)絡(luò)的行人目標(biāo)檢測(cè)方法。并且針對(duì)提取特征訓(xùn)練的分類(lèi)器遍歷搜索慢,在隧道場(chǎng)景下采用運(yùn)動(dòng)信息縮小搜索范圍,節(jié)省了搜索時(shí)間。另外針對(duì)隧道環(huán)境下噪聲大行人特征提取困難的問(wèn)題,利用卷積神經(jīng)網(wǎng)絡(luò)對(duì)特征提取的優(yōu)勢(shì)特點(diǎn),訓(xùn)練了端到端的隧道場(chǎng)景下行人檢測(cè)網(wǎng)絡(luò)。本文的主要研究?jī)?nèi)容如下:(1)一般的行人檢測(cè)分類(lèi)器的訓(xùn)練中通常采用單一的HOG特征,在隧道環(huán)境下檢測(cè)準(zhǔn)確率偏低。本文通過(guò)引入一種局部二值模式(LBP)特征與梯度方向直方圖特征(HOG)串聯(lián)輸入到支持向量機(jī)的分類(lèi)模型中,訓(xùn)練得到的基于聯(lián)合特征行人檢測(cè)器大幅提升了隧道環(huán)境下行人檢測(cè)的準(zhǔn)確率。(2)基于HOG特征與LBP特征串聯(lián)訓(xùn)練的分類(lèi)器一般采用滑動(dòng)窗口遍歷搜索整個(gè)圖像的策略,這樣造成了巨大的時(shí)效性損失。在高速公路隧道中,監(jiān)控畫(huà)面出現(xiàn)在固定場(chǎng)景下。根據(jù)視頻監(jiān)控的這一特點(diǎn),通過(guò)提取行人移動(dòng)信息,將分類(lèi)器檢測(cè)與一種改進(jìn)的高斯混合背景差分方法相結(jié)合,提取圖像中運(yùn)動(dòng)區(qū)域,減少分類(lèi)器對(duì)圖像的搜索次數(shù),大幅提升了算法系統(tǒng)對(duì)行人的識(shí)別效率。(3)針對(duì)高速公路隧道環(huán)境噪聲造成行人與環(huán)境輪廓邊界弱,傳統(tǒng)機(jī)器學(xué)習(xí)方法難以提取有效特征的問(wèn)題,本文利用卷積神經(jīng)網(wǎng)絡(luò)高效的特征提取能力,通過(guò)改進(jìn)候選框提取方法,使用RPN候選框提取網(wǎng)絡(luò),在選用單幅圖片候選框少的情況下訓(xùn)練出行人檢測(cè)的單一目標(biāo)識(shí)別網(wǎng)絡(luò)。對(duì)候選框提取網(wǎng)絡(luò)與行人檢測(cè)網(wǎng)絡(luò)進(jìn)行了訓(xùn)練,得到端到端的行人檢測(cè)網(wǎng)絡(luò)。相對(duì)于特征設(shè)計(jì)的行人檢測(cè)模型,大幅度的提升隧道環(huán)境行人檢測(cè)的準(zhǔn)確率,且在一定程度上提升基于RCNN算法框架下的行人檢測(cè)速度。針對(duì)高速公路隧道環(huán)境下行人檢測(cè)的要求,研究了基于特征提取的分類(lèi)器模型對(duì)隧道應(yīng)用場(chǎng)景的適應(yīng)性,并提出相應(yīng)的方法改進(jìn)檢測(cè)模型。將區(qū)域卷積神經(jīng)網(wǎng)絡(luò)應(yīng)用在氋速公路隧道場(chǎng)景下的行人檢測(cè),同時(shí)訓(xùn)練了端到端的深度行人檢測(cè)模型。提升了隧道監(jiān)控場(chǎng)景下行人檢測(cè)的準(zhǔn)確率,對(duì)基于卷積神經(jīng)網(wǎng)絡(luò)的其他目標(biāo)物識(shí)別工作具有一定的借鑒意義。
[Abstract]:Tunnel area is the key area of highway management. Illegal entry of pedestrians and non-motorized vehicles into expressway tunnel will seriously affect the normal operation of expressway and cause huge safety hazard. The pedestrian detection technology in video surveillance in tunnel environment is an important guarantee for the normal operation of highway. In the tunnel environment, the environment lighting conditions are poor, a lot of noise is produced in the image, the pedestrian in the tunnel has small targets and low pixels. It brings great challenge to pedestrian detection in tunnel environment. This paper mainly studies the segmentation method of foreground target and background object in video detection. The pedestrian target detection method based on mathematical feature extraction and convolution neural network is used. In addition, aiming at the difficult problem of extracting noisy pedestrian features in tunnel environment, we use convolutional neural network to extract features. The downlink detection network of end-to-end tunnel scene is trained. The main contents of this paper are as follows: 1) in the training of pedestrian detection classifier, a single HOG feature is usually used. In this paper, a local binary pattern (LBP) feature and gradient direction histogram feature (hog) are introduced into the classification model of support vector machine in series. The trained pedestrian detector based on joint feature greatly improves the accuracy of pedestrian detection in tunnel environment. (2) the classifier based on HOG feature and LBP feature series training generally uses sliding window traversal strategy to search the whole image. This results in a huge loss of timeliness. In highway tunnels, surveillance images appear in fixed scenes. According to this characteristic of video surveillance, by extracting pedestrian movement information, Combining the classifier detection with an improved Gao Si mixed background differential method, the moving region of the image is extracted, and the search times of the image are reduced. The efficiency of pedestrian recognition in the algorithm system is greatly improved. (3) aiming at the problem that the boundary between pedestrian and environment is weak due to the noise in highway tunnel environment, the traditional machine learning method is difficult to extract effective features. In this paper, we use convolutional neural network to extract features, improve the method of candidate extraction, and use RPN candidate to extract the network. The single target recognition network is trained under the condition of few single image candidate frames, and the candidate extraction network and pedestrian detection network are trained. Get the end to end pedestrian detection network. Compared with the feature designed pedestrian detection model, greatly improve the accuracy of pedestrian detection in tunnel environment. To some extent, the speed of pedestrian detection based on RCNN algorithm is improved. According to the requirement of pedestrian detection in highway tunnel environment, the adaptability of classifier model based on feature extraction to tunnel application scene is studied. The corresponding method is put forward to improve the detection model. The regional convolution neural network is applied to pedestrian detection in the scene of highway tunnel. At the same time, it trains the end-to-end depth pedestrian detection model, improves the accuracy of downlink detection of tunnel monitoring scene, and has some reference significance for other target recognition work based on convolution neural network.
【學(xué)位授予單位】：昆明理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：U458;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 張利平;邵宗凱;吳建德;;基于改進(jìn)KSVD和極限學(xué)習(xí)機(jī)的車(chē)型識(shí)別方法研究[J];計(jì)算機(jī)與數(shù)字工程;2016年06期

2 劉操;鄭宏;黎曦;余典;;基于多通道融合HOG特征的全天候運(yùn)動(dòng)車(chē)輛檢測(cè)方法[J];武漢大學(xué)學(xué)報(bào)(信息科學(xué)版);2015年08期

3 汪成亮;周佳;黃晟;;基于高斯混合模型與PCA-HOG的快速運(yùn)動(dòng)人體檢測(cè)[J];計(jì)算機(jī)應(yīng)用研究;2012年06期

4 王亮,胡衛(wèi)明,譚鐵牛;人運(yùn)動(dòng)的視覺(jué)分析綜述[J];計(jì)算機(jī)學(xué)報(bào);2002年03期

5 徐一華,朱玉文,賈云得;一種人頭部實(shí)時(shí)跟蹤方法[J];中國(guó)圖象圖形學(xué)報(bào);2002年01期

相關(guān)博士學(xué)位論文 前1條

1 趙明;二維視覺(jué)對(duì)象分割[D];浙江大學(xué);2004年

相關(guān)碩士學(xué)位論文 前2條

1 魯寒凝;基于HOG特征的交通標(biāo)志檢測(cè)與識(shí)別算法研究[D];長(zhǎng)安大學(xué);2015年

2 陳虹穎;隧道環(huán)境下行人目標(biāo)視頻檢測(cè)技術(shù)研究[D];重慶大學(xué);2013年

，

本文編號(hào)：1590244