本文選題：無人機 + 聚簇�。� 參考：《廣西大學》2017年碩士論文

【摘要】：輸電線路桿塔的三維重建研究是通過無人機采集輸電線路桿塔的圖像數(shù)據(jù),然后利用計算機視覺技術(shù)進行三維重建,重建出表面致密、紋理豐富、可視化強的三維桿塔結(jié)構(gòu)模型。本研究課題是基于電網(wǎng)的工程應(yīng)用背景而設(shè)立,具體研究內(nèi)容為:基于圖像質(zhì)量約束的無序圖像關(guān)鍵幀提取和采用去抖動模糊算法的稠密三維重建。本課題研究不僅具有理論研究意義而且還具有實際應(yīng)用價值:在理論研究層面,無人機采集的數(shù)字圖像是無序圖像,存在數(shù)目眾多、信息冗余、質(zhì)量參差不齊的問題。無人機采集大場景對象的圖像進行三維重建時因抖動產(chǎn)生的模糊現(xiàn)象以及二維圖像序列經(jīng)運動恢復(fù)結(jié)構(gòu)SFM(Structure From Motion)后得到的三維點云稀疏、可視化差的問題,都是我們急需解決的問題;在實際應(yīng)用層面,無人機以輕便、靈巧、快捷的優(yōu)勢對輸電線路進行巡檢,可以彌補人工目測有死角、盲點的缺點,同時可以進入地形復(fù)雜、環(huán)境惡劣、巡檢人員不宜進入的地方巡檢,巡檢效率高、安全系數(shù)高。巡檢員可不去桿塔處去目測,直接在電腦面前對采集的桿塔圖像信息進行立體視覺三維重建,使重建的三維桿塔效果逼真,細節(jié)豐富,可以滿足巡檢的視覺要求,本課題將有效的節(jié)省人力、物力,為電網(wǎng)的安全運行保駕護航。本課題包含基于圖像質(zhì)量約束的無序圖像關(guān)鍵幀提取和采用去抖動模糊算法的稠密三維重建。具體研究內(nèi)容如下:(1)本課題根據(jù)無人機巡檢、圖像采集的需求,建立四旋翼無人機圖像采集系統(tǒng)。根據(jù)各子系統(tǒng)的結(jié)構(gòu)功能對器件進行選型,搭建無人機輸電線路圖像采集平臺,測試、驗證平臺的穩(wěn)定性。(2)分析、對比無序圖像的聚簇和關(guān)鍵幀提取算法:本課題首先采用不預(yù)設(shè)K-均值的聚簇算法對無序圖像進行自動聚簇;其次根據(jù)相似距離從每簇中提取出離聚簇中心最近的一幀作為關(guān)鍵幀;最后采用二次模糊處理算法對提取的關(guān)鍵幀進行質(zhì)量評價,其評價值若滿足質(zhì)量要求則保留,不滿足則返回原來的簇中重新進行關(guān)鍵幀的提取與評價,直到提取的關(guān)鍵幀滿足質(zhì)量要求為止。(3)分析、對比航拍圖像去抖動模糊算法和三維重建算法:本文首先采用去抖動模糊算法恢復(fù)模糊圖像的原始圖像信息,然后在運動恢復(fù)結(jié)構(gòu)的基礎(chǔ)上進行基于點云的稠密三維重建,最后對稠密重建后的點云進行泊松表面重建以得到表面致密、均勻、色彩豐富的三維桿塔模型。綜上所述:基于無人機航拍輸電線路桿塔的三維重建課題,提出基于圖像質(zhì)量約束的無序圖像關(guān)鍵幀提取方法,提取出質(zhì)量高、信息豐富的關(guān)鍵幀。提出采用去抖動模糊的稠密三維重建算法,重建出表面致密、紋理豐富、效果逼真的三維結(jié)構(gòu)模型,因此本課題兼具理論研究意義與實際應(yīng)用價值。
[Abstract]:The research of 3D reconstruction of transmission line tower is to collect the image data of transmission line tower by UAV, and then use computer vision technology to reconstruct the three dimensional transmission line tower. The reconstructed surface is compact and rich in texture. Strong visualization of three-dimensional tower structure model. This research is based on the engineering application background of power grid. The research contents are as follows: key frame extraction of disordered image based on image quality constraint and dense 3D reconstruction using jitter removing fuzzy algorithm. The research of this subject is not only of theoretical significance but also of practical application value: at the level of theoretical research, the digital images collected by UAV are disordered images, there are many problems, such as large number, information redundancy and uneven quality. The fuzzy phenomenon caused by jitter and the problem of sparse point cloud and poor visualization of 2D image sequence obtained by SFM (structure from Motion), which are used for 3D reconstruction of large scene object by UAV, are discussed in this paper. At the practical application level, unmanned aerial vehicles patrol transmission lines with the advantages of portability, dexterity, and speed, which can make up for the shortcomings of manual visual measurement of dead corners and blind spots, and at the same time, they can enter the complex terrain. The environment is abominable, the patrol personnel should not enter the place to patrol, the patrol inspection efficiency is high, the safety factor is high. Instead of going to the pole tower for visual measurement, the inspector can directly carry out three-dimensional visual reconstruction of the collected tower image information in front of the computer, making the reconstructed three-dimensional tower effect realistic and rich in detail, which can meet the visual requirements of the inspection tour. This subject will save manpower and material resources effectively, and guarantee the safe operation of power grid. This thesis includes key frame extraction based on image quality constraint and dense 3D reconstruction using jitter-removing fuzzy algorithm. The specific research contents are as follows: (1) according to the requirements of UAV inspection and image acquisition, a four-rotor UAV image acquisition system is established. According to the structure and function of each subsystem, the device is selected, the UAV transmission line image acquisition platform is set up, the stability of the platform is tested and verified. (2) Analysis, Compare the clustering algorithm of unordered image and the key frame extraction algorithm: firstly, we use the clustering algorithm without presupposing K- mean to cluster the unordered image automatically; Secondly, according to the similarity distance, the nearest frame of the cluster center is extracted from each cluster as the key frame. Finally, the second fuzzy algorithm is used to evaluate the quality of the extracted key frame, and the evaluation value is retained if the quality requirement is satisfied. If the key frame is not satisfied, the key frame will be extracted and evaluated again in the original cluster until the extracted key frame meets the quality requirement. (3) Analysis, Compared with the dejitter blur algorithm and 3D reconstruction algorithm of aerial image: firstly, the original image information of blur image is restored by de-jitter blur algorithm, and then dense 3D reconstruction based on point cloud is carried out on the basis of motion recovery structure. Finally, Poisson surface reconstruction is carried out on the dense point cloud to obtain a dense, uniform and colorful three-dimensional tower model. To sum up: based on the three-dimensional reconstruction of aerial photography transmission line tower, an image quality constraint based unordered image key frame extraction method is proposed to extract high-quality and informative key frames. A dense 3D reconstruction algorithm with dejitter ambiguity is proposed to reconstruct 3D structure model with dense surface rich texture and realistic effect. Therefore this paper has both theoretical research and practical application value.
【學位授予單位】：廣西大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP391.41;TM754

【參考文獻】

相關(guān)期刊論文 前10條

1 田金沙;韓永國;吳亞東;趙小樂;張紅英;;基于尺度不變性的無參考圖像質(zhì)量評價[J];計算機應(yīng)用;2016年03期

2 王威;劉婧;楊蔚蔚;李驥;;基于DCT域紋理結(jié)構(gòu)相似度的模糊圖像質(zhì)量評價[J];計算機工程;2015年11期

3 何健;羅f3昀;范暢;侯品;;架空輸電線路無人機人工聯(lián)合巡檢模式初探[J];電子測試;2015年21期

4 張平;王山東;黃瑾娉;周明明;;基于SFM和CMVS/PMVS的建筑物點云重構(gòu)方法研究[J];蘇州科技學院學報(自然科學版);2015年03期

5 李勇志;支曉棟;唐海龍;李丹;趙政;孫長奎;;無人機遙感技術(shù)在農(nóng)業(yè)中的發(fā)展與應(yīng)用[J];安徽農(nóng)業(yè)科學;2015年25期

6 張淑芳;張聰;張濤;雷志春;;通用型無參考圖像質(zhì)量評價算法綜述[J];計算機工程與應(yīng)用;2015年19期

7 李濱;王宏宇;楊笑天;;無人機系統(tǒng)在森林防火方面的應(yīng)用及其發(fā)展[J];電子技術(shù);2015年05期

8 李俊峰;;基于RGB色彩空間自然場景統(tǒng)計的無參考圖像質(zhì)量評價[J];自動化學報;2015年09期

9 彭向陽;劉正軍;麥曉明;羅智斌;王柯;謝小偉;;無人機電力線路安全巡檢系統(tǒng)及關(guān)鍵技術(shù)[J];遙感信息;2015年01期

10 李順意;侯進;甘凌云;;基于幀間距的運動關(guān)鍵幀提取[J];計算機工程;2015年02期

相關(guān)碩士學位論文 前1條

1 王一;基于無人直升機航拍照片的三維目標重建技術(shù)[D];上海交通大學;2010年

，

本文編號：2081016