本文選題：快速視頻拼接 + 各向同性Sobel算子�。� 參考：《內(nèi)蒙古農(nóng)業(yè)大學(xué)》2017年碩士論文

【摘要】：由于受自身廣角限制,單一攝像頭無法得到一系列清晰度較高的視頻圖像。為了更好地去解決單一攝像頭的缺陷,視頻拼接技術(shù)因此得到發(fā)展。伴跟著計算機(jī)技術(shù)深入研究,尤其是利用計算機(jī)視覺的行車智能系統(tǒng)、智能車載技術(shù)和主動行駛的研究,實時捕獲到目標(biāo)物體360°范圍內(nèi)的視覺圖像,也正被各個領(lǐng)域的研究學(xué)者不斷的關(guān)注。本文針對快速視頻拼接算法的拼接性能以及拼接所呈現(xiàn)出的效果進(jìn)行了研究,重點在一下幾個方面展開工作:(1)本文采用普通的CCD攝像頭組成視頻采集單元,多路攝像頭采集通道的同步問題進(jìn)行了研究。同時,本文對輸入視頻圖像進(jìn)行預(yù)先處理,消除因為外部光線以及視角問題產(chǎn)生的圖像質(zhì)量不高以及攝像頭本身產(chǎn)生畸變等問題,有效的優(yōu)化了拼接質(zhì)量。(2)本文分析了基于傳統(tǒng)的Harris角點算法以及SIFT特征算法的配準(zhǔn)方法�；赟IFT特征的算法運行準(zhǔn)確度高,但實驗流程較為復(fù)雜,無法滿足快速視頻拼接;基于Harris角點的配準(zhǔn)算法,復(fù)雜度較低,實驗運行時間短。最后本文決定采用基于Harris角點的拼接算法,并對原始Harris存在特定點檢測準(zhǔn)確率低、運行過程復(fù)雜等問題,進(jìn)行了改進(jìn)和優(yōu)化,在不降低拼接質(zhì)量的同時,加快了速度。(3)此外,在視頻融合方面,本文采用兩步處理的方法,通過分布處理,結(jié)合線性融合的方法,得到了可以應(yīng)用于本文實驗的融合方法,實現(xiàn)了快速視頻拼接。作為視頻拼接的重點,視頻幀圖像的配對以及融合起到十分核心的地位,本文針對原始Harris角點算法進(jìn)行改進(jìn),采用更合理的各向同性Sobel算子對角點進(jìn)行處理,同時,利用四鄰域判別法,快速選擇特征點,加快了算法運行速率。針對算法改進(jìn),本文進(jìn)行了多組實驗,最終結(jié)果表明本文提出的改進(jìn)算法是切實有效的。
[Abstract]:Because of its wide-angle limitation, a single camera can not get a series of high-definition video images. In order to solve the defect of single camera, video stitching technology has been developed. Following the in-depth study of computer technology, especially the intelligent driving system using computer vision, intelligent vehicle technology and active driving research, the visual images of the target object are captured in real time in the range of 360 擄. It is also being paid more and more attention by researchers in various fields. In this paper, the performance of fast video stitching algorithm and the effect of stitching are studied. The emphasis is on the following several aspects: 1) this paper uses the common CCD camera to form the video capture unit. The synchronization of multi-channel camera acquisition channels is studied. At the same time, the input video image is pre-processed in this paper to eliminate the problems of poor image quality caused by external light and angle of view, and distortion of the camera itself. This paper analyzes the registration method based on traditional Harris corner algorithm and SIFT feature algorithm. The algorithm based on SIFT features has high running accuracy, but the experimental flow is more complex, which can not meet the fast video stitching. The registration algorithm based on Harris corner has low complexity and short running time. Finally, this paper adopts the algorithm based on Harris corner, and improves and optimizes the original Harris, which has some problems, such as low detection accuracy of specific points and complex running process, which accelerates the speed of stitching without reducing the quality of stitching. In the aspect of video fusion, this paper adopts the two-step processing method, through the distributed processing, combining the linear fusion method, obtains the fusion method which can be applied to the experiment in this paper, and realizes the fast video stitching. As the focus of video mosaic, the pairing and fusion of video frame images play a very important role. This paper improves the original Harris corner algorithm, uses more reasonable isotropic Sobel operator to deal with the corner points, at the same time, The fast selection of feature points by using the four neighborhood discriminant method speeds up the running speed of the algorithm. Several experiments are carried out to improve the algorithm. The final results show that the improved algorithm is effective and effective.
【學(xué)位授予單位】：內(nèi)蒙古農(nóng)業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 陳澤紅;黃靈操;;基于SIFT算法的圖像匹配研究[J];河北北方學(xué)院學(xué)報(自然科學(xué)版);2014年04期

2 王小強(qiáng);陳臨強(qiáng);梁旭;;實時全自動視頻拼接方法[J];計算機(jī)工程;2011年05期

3 于麗莉;戴青;;一種改進(jìn)的SIFT特征匹配算法[J];計算機(jī)工程;2011年02期

4 曲天偉;安波;陳桂蘭;;改進(jìn)的RANSAC算法在圖像配準(zhǔn)中的應(yīng)用[J];計算機(jī)應(yīng)用;2010年07期

5 毛雁明;蘭美輝;王運瓊;馮喬生;;一種改進(jìn)的基于Harris的角點檢測方法[J];計算機(jī)技術(shù)與發(fā)展;2009年05期

6 趙萬金;龔聲蓉;劉純平;沈項軍;;一種自適應(yīng)的Harris角點檢測算法[J];計算機(jī)工程;2008年10期

7 朱云芳;葉秀清;顧偉康;;視頻序列的全景圖拼接技術(shù)[J];中國圖象圖形學(xué)報;2006年08期

8 張光年;;多尺度小波梯度圖像增強(qiáng)方法[J];首都師范大學(xué)學(xué)報(自然科學(xué)版);2006年02期

9 封靖波,蘇志勛,劉秀平;一種基于相似曲線的全景圖自動拼接算法[J];計算機(jī)學(xué)報;2003年11期

10 雷成,胡占義,吳福朝,TSUI H T;一種新的基于Kruppa方程的攝像機(jī)自標(biāo)定方法[J];計算機(jī)學(xué)報;2003年05期

相關(guān)碩士學(xué)位論文 前2條

1 潘子昂;基于SIFT算法的圖像匹配研究[D];西安電子科技大學(xué);2012年

2 江舒;基于角點的圖像配準(zhǔn)問題研究[D];上海交通大學(xué);2008年

，

本文編號：1887763