發(fā)布時(shí)間：2018-05-06 18:18

  本文選題：時(shí)間戳移除 + 時(shí)間戳重置��； 參考：《華中師范大學(xué)》2016年博士論文

【摘要】：多相機(jī)視頻融合是許多智能化教育系統(tǒng)中的一個(gè)重要子系統(tǒng)。校園大型場館的監(jiān)控,教室內(nèi)全景視頻的錄制,教育機(jī)器人多視角的視覺感知等系統(tǒng)都需要多個(gè)攝像機(jī)同步采集視頻并合成全景視頻。視頻中時(shí)間戳的移除與重置是這個(gè)全景生成過程中的一個(gè)有挑戰(zhàn)性的研究問題。當(dāng)每個(gè)視頻中有一個(gè)疊加的時(shí)間戳而又沒有被移除的情況下,由多個(gè)視頻合成的全景視頻會(huì)同時(shí)出現(xiàn)多個(gè)時(shí)間戳。這些時(shí)間戳影響視覺效果,甚至?xí)趽踝∫恍└信d趣的對(duì)象。為了產(chǎn)生一個(gè)高質(zhì)量的全景視頻,這些時(shí)間戳需被移除,在合成的全景視頻中也需要重新疊加一個(gè)時(shí)間戳。本文研究問題是如何高效地移除單個(gè)視頻中出現(xiàn)的時(shí)間戳,并在全景視頻中高質(zhì)量地重置時(shí)間戳。其主要目標(biāo)是研究全景視頻融合功能里面的時(shí)間戳移除與重置過程中所產(chǎn)生的一系列問題的解決方法和關(guān)鍵技術(shù),并實(shí)現(xiàn)這些提出的方法在教育系統(tǒng)中的應(yīng)用。其研究內(nèi)容包括六個(gè)方面:(1)時(shí)間戳數(shù)字定位;(2)時(shí)間戳區(qū)域移除與修復(fù);(3)多相機(jī)顏色校正;(4)時(shí)間戳數(shù)字識(shí)別;(5)全景視頻生成;(6)教育機(jī)器人全景視頻系統(tǒng)的實(shí)現(xiàn)。通過以上六個(gè)方面內(nèi)容的研究,本論文在學(xué)術(shù)上做出了一些貢獻(xiàn),其主要貢獻(xiàn)有如下三點(diǎn):(1)在時(shí)間戳數(shù)字區(qū)域移除與修復(fù)問題上,傳統(tǒng)的圖像修復(fù)方法一般是利用近鄰區(qū)域的像素或者紋理來填充,無法還原出真實(shí)圖像。針對(duì)這個(gè)難點(diǎn)問題,本文提出通過主動(dòng)控制PTZ相機(jī)使其周期性運(yùn)動(dòng)移位,使被時(shí)間戳數(shù)字遮住的區(qū)域能夠在時(shí)間上鄰近的幀中顯露出來,然后通過快速圖像平移量估計(jì)算法和相鄰幀像素恢復(fù)方法來實(shí)現(xiàn)時(shí)間戳數(shù)字區(qū)域的移除與修復(fù)�？焖賵D像平移量估計(jì)算法采用PTZ運(yùn)動(dòng)估計(jì)結(jié)合全局直方圖匹配來計(jì)算移位兩幀之間的平移量,該方法計(jì)算量小從而使其有較好的實(shí)時(shí)性。相鄰幀像素恢復(fù)方法是根據(jù)快速圖像平移量估計(jì)算法求得的結(jié)果,來定位原來被時(shí)間戳數(shù)字遮擋的區(qū)域在相鄰幀中的位置,并利用圖像融合技術(shù)來修復(fù)原來視頻幀中的時(shí)間戳區(qū)域。采用該方法移除時(shí)間戳所得到的視頻具有高度的場景一致性。(2)在多相機(jī)視頻顏色校正問題上,提出一個(gè)混合的顏色校正技術(shù),該技術(shù)將區(qū)域匹配方法與伽瑪校正和線性校正方法相結(jié)合,解決了相鄰相機(jī)之間重疊區(qū)域不規(guī)則情況下的顏色校正問題。采用該技術(shù)進(jìn)行顏色校正,即使重疊區(qū)域不規(guī)則,目標(biāo)圖像的色彩也能與參考圖像保持高度一致。(3)在搭建的教育機(jī)器人平臺(tái)上,設(shè)計(jì)并實(shí)現(xiàn)了全景視頻系統(tǒng)。該系統(tǒng)實(shí)現(xiàn)了時(shí)間戳移除與重置功能。
[Abstract]:Multi-camera video fusion is an important subsystem in many intelligent education systems. The monitoring of large campus venues, the recording of panoramic video in the classroom, the visual perception of educational robots and so on require multiple cameras to synchronize the acquisition of video and synthesis of panoramic video. The removal and reset of time stamps in video is a challenging problem in the panorama generation process. When each video has a superimposed timestamp without being removed, the panoramic video synthesized by multiple videos will have multiple timestamps at the same time. These timestamps affect visual effects and may even block objects of interest. In order to produce a high quality panoramic video, these timestamps need to be removed, and a new timestamp is also needed to be overlaid in the synthesized panoramic video. In this paper, the problem is how to remove the timestamp in a single video efficiently and reset the timestamp in the panoramic video with high quality. Its main goal is to study the solutions and key technologies of a series of problems in the process of removing and resetting timestamp in the panoramic video fusion function, and to realize the application of these proposed methods in the education system. The research contents include six aspects: 1) time stamp digital location / 2) time stamp area removal and repair (3) multi-camera color correction / 4) time stamp digital recognition / 5) panoramic video generation / 6) implementation of a panoramic video system for educational robots. Through the research of the above six aspects, this paper has made some academic contributions, the main contributions are as follows: 3 points: 1) on the removal and repair of time stamp digital area, The traditional image restoration method is usually filled with pixels or textures of adjacent regions, which can not restore the real image. Aiming at this difficult problem, this paper proposes that the PTZ camera can be moved periodically by actively controlling it, so that the region hidden by the timestamp number can be exposed in the frame adjacent to the time. Then the fast image translation estimation algorithm and the adjacent frame pixel recovery method are used to realize the removal and restoration of the timestamp digital region. Fast image translation estimation algorithm uses PTZ motion estimation combined with global histogram matching to calculate the translation between two shifting frames. The pixel restoration method of adjacent frames is based on the result of fast image translation estimation algorithm to locate the position of the previously digitally occluded region in adjacent frames. The image fusion technique is used to repair the time stamp region in the original video frame. Using this method to remove the timestamp, the video has a high scene consistency. (2) A hybrid color correction technique is proposed for multi-camera video color correction. This technique combines region matching with gamma correction and linear correction to solve the problem of color correction in the case of irregular overlapped regions between adjacent cameras. Even if the overlapping area is irregular, the color of the target image can be highly consistent with the reference image. (3) on the platform of the educational robot, a panoramic video system is designed and implemented. The system realizes the function of time stamp removal and reset.
【學(xué)位授予單位】：華中師范大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：G434

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 許宏科;秦嚴(yán)嚴(yán);陳會(huì)茹;;基于改進(jìn)ORB的圖像特征點(diǎn)匹配[J];科學(xué)技術(shù)與工程;2014年18期

2 何賓;陶丹;彭勃;;高實(shí)時(shí)性F-SIFT圖像拼接算法[J];紅外與激光工程;2013年S2期

3 曾丹;陳劍;張琦;史浩;;基于全局拓?fù)浣Y(jié)構(gòu)的分級(jí)三角剖分圖像拼接[J];計(jì)算機(jī)研究與發(fā)展;2012年01期

4 徐榮國;黃力;;改進(jìn)的基于SVM的車輛牌照識(shí)別技術(shù)[J];廣西民族大學(xué)學(xué)報(bào)(自然科學(xué)版);2010年04期

5 李文宏;田文娟;王霞;駱科學(xué);;基于支持向量機(jī)的人民幣紙幣序列號(hào)識(shí)別方法[J];信息與控制;2010年04期

6 Kenji Yamamoto;Ryutaro Oi;;Color Correction for Multi-view Video Using Energy Minimization of View Networks[J];International Journal of Automation & Computing;2008年03期

7 李興陽;王文青;;基于模板匹配的車牌字符識(shí)別算法研究[J];陜西師范大學(xué)學(xué)報(bào)(自然科學(xué)版);2008年S1期

8 王欣;殷肖川;周翔翔;;一種改進(jìn)的模版匹配識(shí)別算法[J];計(jì)算機(jī)工程與應(yīng)用;2007年36期

9 李曉明;趙訓(xùn)坡;鄭鏈;胡占義;;基于Fourier-Mellin變換的圖像配準(zhǔn)方法及應(yīng)用拓展[J];計(jì)算機(jī)學(xué)報(bào);2006年03期

10 張二虎,卞正中;基于最大熵和互信息最大化的特征點(diǎn)配準(zhǔn)算法[J];計(jì)算機(jī)研究與發(fā)展;2004年07期

相關(guān)碩士學(xué)位論文 前3條

1 楊瑞陽;基于SURF算法的醫(yī)學(xué)顯微圖像拼接研究[D];蘭州大學(xué);2014年

2 龔軍輝;基于改進(jìn)SURF的X線醫(yī)學(xué)圖像拼接算法研究[D];云南大學(xué);2013年

3 張亞娟;基于SURF特征的圖像與視頻拼接技術(shù)的研究[D];西安電子科技大學(xué);2013年

，

本文編號(hào)：1853360