【摘要】：由于無人機(jī)航測(cè)遙感系統(tǒng)具有靈活、成本低、大比例尺高精度的特點(diǎn)，在小區(qū)域和飛行困難地區(qū)快速獲取高分辨率影像方面有明顯優(yōu)勢(shì)。因此，無人機(jī)航測(cè)遙感技術(shù)已經(jīng)成為提高測(cè)繪成果現(xiàn)勢(shì)性的有力手段，是增強(qiáng)測(cè)繪應(yīng)急保障能力的捷徑。無人機(jī)影像由于受飛行高度、相機(jī)視角的影響，單張無人機(jī)影像所覆蓋的區(qū)域面積不大，在特定任務(wù)中需要對(duì)多張影像進(jìn)行拼接，有效覆蓋所有工作區(qū)。影像匹配從提出到現(xiàn)在，經(jīng)過了無數(shù)次的改進(jìn)和發(fā)展，無論是匹配點(diǎn)精度還是匹配速度都有了質(zhì)和量的飛躍，但是由于無人機(jī)影像具有像幅小、重疊度變化大、旋偏角大、影像畸變大、噪聲和遮擋嚴(yán)重等特性，需要尋找一種對(duì)各種畸變、噪聲都具有良好魯棒性的一種算法，而基于特征匹配的影像拼接算法能很好的滿足，因而被廣泛的應(yīng)用。本文主要以基于尺度不變特征（SFIT）的影像匹配算法為主要內(nèi)容，學(xué)習(xí)無人機(jī)影像快速自動(dòng)拼接的關(guān)鍵技術(shù)，并利用C++編程實(shí)現(xiàn)，，論文的研究工作主要包括以下幾個(gè)部分： 1.總結(jié)了無人機(jī)影像拼接技術(shù)的意義和國(guó)內(nèi)外研究現(xiàn)狀，確定了本文無人機(jī)影像拼接的技術(shù)流程。 2.利用C++語言編程，VS2010平臺(tái)編譯實(shí)現(xiàn)了無地理坐標(biāo)的無人機(jī)遙感影像快速自動(dòng)拼接。采用SIFT算法進(jìn)行影像特征點(diǎn)提取、同名點(diǎn)匹配，然后利用RANSAC剔除誤匹配點(diǎn)對(duì)，完成精匹配。最后利用同名點(diǎn)對(duì)解算影像間的幾何變換模型，完成一條航帶多張影像拼接，輸出成果。 3.采用直接加權(quán)平均法對(duì)拼接影像進(jìn)行融合，處理色差、光照差異、拼接縫等問題。 研究結(jié)果表明，采用SIFT算法能夠有效提取大量的特征點(diǎn)用于影像匹配，對(duì)缺乏地面控制點(diǎn)的無人機(jī)影像拼接效果良好。但是由于特征點(diǎn)過多會(huì)影響計(jì)算速度，需要尋求有效方法進(jìn)行過濾，更好的滿足無人機(jī)遙感的時(shí)效性要求。
[Abstract]:Because UAV aerial survey remote sensing system has the characteristics of flexibility, low cost and high precision on large scale, it has obvious advantages in obtaining high-resolution images quickly in small area and difficult area of flight. Therefore, UAV aerial remote sensing technology has become a powerful means to improve the present situation of surveying and mapping achievements, and it is a shortcut to enhance the ability of emergency support of surveying and mapping. Due to the influence of flying altitude and camera angle, the area covered by single UAV image is not large, so it is necessary to splice multiple images to cover all the working areas effectively in a given task. Since it was put forward, image matching has been improved and developed countless times. Both the precision of matching point and the speed of matching have made a leap in quality and quantity. However, due to the small image size, large overlap and large rotation angle, the UAV image has a small image amplitude, a large degree of overlap, and a large rotation angle. Because of its large distortion, serious noise and occlusion, it is necessary to find an algorithm that has good robustness to all kinds of distortion and noise. However, the image mosaic algorithm based on feature matching can be satisfied very well, so it has been widely used. In this paper, the image matching algorithm based on scale invariant feature (SFIT) is taken as the main content, and the key technology of fast automatic stitching of UAV image is studied, and realized by C programming. The research work of this paper mainly includes the following parts: 1. This paper summarizes the significance of UAV image mosaic technology and the research status at home and abroad, and determines the technical flow of UAV image splicing in this paper. 2. By programming in C language and compiling on VS2010 platform, the remote sensing image of unmanned aerial vehicle (UAV) without geographical coordinates can be automatically stitched quickly and automatically. The SIFT algorithm is used to extract the feature points and match the same name points. Then the mismatched point pairs are eliminated by RANSAC to complete the fine matching. Finally, using the geometric transformation model of the same-named point pair to solve the image, we complete the multi-image splicing of one airstrip and output the result. 3. The direct weighted average method is used to fuse the splicing image, deal with the color difference, illumination difference, splicing seam and so on. The results show that the SIFT algorithm can effectively extract a large number of feature points for image matching and has a good effect on UAV image mosaic without ground control points. However, because too many feature points will affect the computing speed, it is necessary to find an effective method to filter and better meet the requirements of the time-effectiveness of UAV remote sensing.
【學(xué)位授予單位】：中國(guó)地質(zhì)大學(xué)（北京）
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：P231;P237

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 李原福;王學(xué)明;段金輝;;基于特征點(diǎn)檢測(cè)的重疊圖像拼接算法[J];吉林大學(xué)學(xué)報(bào)(信息科學(xué)版);2010年06期

2 雷小群;李芳芳;肖本林;;一種基于改進(jìn)SIFT算法的遙感影像配準(zhǔn)方法[J];測(cè)繪科學(xué);2010年03期

3 邱建國(guó);張建國(guó);李凱;;基于Harris與Sift算法的圖像匹配方法[J];測(cè)試技術(shù)學(xué)報(bào);2009年03期

4 陳裕;劉慶元;;基于SIFT算法和馬氏距離的無人機(jī)遙感圖像配準(zhǔn)[J];測(cè)繪與空間地理信息;2009年06期

5 李波;一種基于小波和區(qū)域的圖像拼接方法[J];電子科技;2005年04期

6 徐光著;朱冰蓮;豐建軍;;一種改進(jìn)的動(dòng)態(tài)場(chǎng)景拼接算法[J];電子科技;2011年07期

7 衛(wèi)征;方俊永;張兵;;非量測(cè)相機(jī)鏡頭光學(xué)畸變的改正[J];光學(xué)技術(shù);2007年06期

8 宮本旭;李亮;;無人機(jī)遙感數(shù)據(jù)的獲取和在礦山監(jiān)測(cè)中的處理方法[J];貴州地質(zhì);2011年03期

9 程紅;陳文劍;;基于SIFT算法的圖像匹配剔點(diǎn)方法研究[J];地理與地理信息科學(xué);2012年06期

10 馬超;趙西安;王青松;;基于均勻特征匹配的無人機(jī)影像拼接[J];北京建筑工程學(xué)院學(xué)報(bào);2013年04期

本文編號(hào)：2460289