本文選題：云導(dǎo)風(fēng) + 局部特征提取��； 參考：《山東大學(xué)》2017年碩士論文

【摘要】：氣象數(shù)值預(yù)報(bào),是一個(gè)與科技民生息息相關(guān)的領(lǐng)域,隨著氣象衛(wèi)星技術(shù)的不斷進(jìn)步,得到的資料也越來(lái)越豐富,氣象數(shù)值預(yù)報(bào)也從依賴專業(yè)人員的經(jīng)驗(yàn),轉(zhuǎn)而利用各類圖像處理技術(shù),使結(jié)果更加客觀,更加精準(zhǔn)。云導(dǎo)風(fēng)技術(shù)主要指利用連續(xù)時(shí)序的衛(wèi)星云圖,捕捉云圖上示蹤云團(tuán)的運(yùn)動(dòng)軌跡以反演風(fēng)場(chǎng)的過(guò)程,是氣象數(shù)值預(yù)報(bào)的一個(gè)重要信息來(lái)源,尤其是對(duì)于較為偏遠(yuǎn)且氣象監(jiān)控站點(diǎn)稀缺的地區(qū)。目前主流方法是基于相關(guān)度計(jì)算的模板匹配方法。然而云團(tuán)的運(yùn)動(dòng)屬于非剛性物體的半流體運(yùn)動(dòng),目前來(lái)講,并沒(méi)有一個(gè)相對(duì)成熟的計(jì)算模型,現(xiàn)有的計(jì)算方法主要利用區(qū)域匹配,對(duì)模板在搜索區(qū)域內(nèi)進(jìn)行相關(guān)度計(jì)算得到云團(tuán)位移,并沒(méi)有充分考慮云團(tuán)特征,對(duì)于有旋風(fēng)帶來(lái)的云團(tuán)旋轉(zhuǎn)并不能夠很好的適應(yīng),并存在搜索方式計(jì)算量較大、對(duì)人工依然有很大依賴等缺點(diǎn),因此該方向依然有很大的挖掘空間和探索價(jià)值。本文為了更好地捕捉云團(tuán)運(yùn)動(dòng)規(guī)律,提高云導(dǎo)風(fēng)計(jì)算效率,考慮將SIFT算法得到的特征點(diǎn)在前后云圖上的分布規(guī)律作為云團(tuán)局部特征信息,并結(jié)合到模板匹配中,是對(duì)云導(dǎo)風(fēng)技術(shù)的一個(gè)新的嘗試。SIFT算法可以有效克服尺度變化、旋轉(zhuǎn)、亮度變化等影響,因此,由SIFT算法產(chǎn)生的特征點(diǎn)可以代表云團(tuán)運(yùn)動(dòng)并對(duì)云團(tuán)局部特征進(jìn)行描述。在本文中,為了避免由于云團(tuán)不確定運(yùn)動(dòng)帶來(lái)的干擾,本算法沒(méi)有直接選用特征點(diǎn)的匹配距離作為移動(dòng)矢量,而是以特征點(diǎn)分布情況作為云團(tuán)的局部特征代表,將模板匹配與局部特征結(jié)合,一方面可以避免對(duì)于云團(tuán)局部存在平滑區(qū)域所造成的特征點(diǎn)難以檢測(cè),另一方面利用局部特征區(qū)域直接匹配,有效縮減模板匹配時(shí)遍歷搜索過(guò)程的計(jì)算量。實(shí)驗(yàn)結(jié)果表明,基于局部特征的區(qū)域匹配算法能夠適應(yīng)云團(tuán)的各類變化,也能夠基于特征點(diǎn)分布得到一定的有旋風(fēng)旋轉(zhuǎn)信息,并且相較于傳統(tǒng)算法能夠獲取更多的風(fēng)矢量信息。尤其對(duì)于高分辨率的衛(wèi)星圖像,在保證一定準(zhǔn)確性的同時(shí),能夠使計(jì)算效率大幅提升。
[Abstract]:The meteorological numerical forecast is a field closely related to the people's livelihood of science and technology. With the continuous progress of meteorological satellite technology, the data obtained are more and more abundant. The weather numerical forecast also depends on the experience of professionals. Instead, various image processing techniques are used to make the results more objective and accurate. Cloud wind guide technology mainly refers to the process of retrieving the wind field by capturing the track track of the cloud cluster on the continuous time series satellite cloud image, which is an important source of information for the meteorological numerical forecast. Especially for the more remote and scarce meteorological monitoring sites. At present, the main method is template matching based on correlation calculation. However, the motion of cloud cluster belongs to the semi-fluid motion of non-rigid object. At present, there is not a relatively mature calculation model. When calculating the correlation degree of the template in the search area, we can get the cloud cluster displacement without fully considering the cloud cluster characteristics, which can not be well adapted to the whirlwind cloud rotation, and there is a large amount of calculation in the search mode. There is still a great dependence on labor, so this direction still has great space and exploration value. In this paper, in order to better capture the cloud motion law and improve the efficiency of cloud wind guide calculation, we consider the distribution of the feature points on the front and rear cloud images obtained by SIFT algorithm as the local feature information of the cloud cluster, and combine it with template matching. SIFT algorithm can effectively overcome the influence of scale change, rotation, brightness change and so on. Therefore, the feature points generated by SIFT algorithm can represent the cloud movement and describe the local features of the cloud cluster. In this paper, in order to avoid the interference caused by the uncertain motion of the cloud, the matching distance of the feature points is not directly selected as the moving vector, but the distribution of the feature points is taken as the local feature of the cloud cluster. Combining template matching with local feature, on the one hand, it can avoid the difficulty of detecting feature points caused by the existence of smooth region in cloud cluster, on the other hand, it can use local feature region to match directly. Effectively reduces the computational cost of traversing the search process when template matching is performed. The experimental results show that the region matching algorithm based on local features can adapt to all kinds of changes of cloud clusters, and can obtain some whirlwind rotation information based on the distribution of feature points, and can obtain more wind vector information than the traditional algorithm. Especially for high resolution satellite images, the computational efficiency can be greatly improved while ensuring certain accuracy.
【學(xué)位授予單位】：山東大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：P456.7;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 萬(wàn)曉敏;田偉紅;韓威;王瑞文;張其松;張曉虎;;FY-2E云導(dǎo)風(fēng)的算法改進(jìn)及其在GRAPES中的同化應(yīng)用研究[J];氣象;2017年01期

2 陳劍虹;韓小珍;;結(jié)合FAST-SURF和改進(jìn)k-d樹最近鄰查找的圖像配準(zhǔn)[J];西安理工大學(xué)學(xué)報(bào);2016年02期

3 董瑤海;;風(fēng)云四號(hào)氣象衛(wèi)星及其應(yīng)用展望[J];上海航天;2016年02期

4 胡逢超;;靜觀風(fēng)雨 感知冷暖——“風(fēng)云家族”的作為與貢獻(xiàn)[J];太空探索;2015年02期

5 馬俠霖;羅鵬;陳志斌;蔡銘;;基于尺度不變特征變換的云導(dǎo)風(fēng)計(jì)算方法[J];氣象科技;2014年03期

6 楊平呂;劉祥余;郭付才;彭新偉;黃思訓(xùn);項(xiàng)杰;;云導(dǎo)風(fēng)的初步探究[J];數(shù)學(xué)的實(shí)踐與認(rèn)識(shí);2013年15期

7 邵壽平;韓春燕;謝勇;琚生根;;改進(jìn)序貫相似性檢測(cè)算法的遙感圖像匹配[J];四川大學(xué)學(xué)報(bào)(自然科學(xué)版);2013年02期

8 趙世軍;蘇小勇;孫學(xué)金;劉磊;高太長(zhǎng);;地基紅外云圖云導(dǎo)風(fēng)方法研究進(jìn)展[J];大氣與環(huán)境光學(xué)學(xué)報(bào);2012年06期

9 朱志文;沈占鋒;駱劍承;;改進(jìn)SIFT點(diǎn)特征的并行遙感影像配準(zhǔn)[J];遙感學(xué)報(bào);2011年05期

10 ;叱咤風(fēng)云的哼哈二將——“風(fēng)云”系列氣象衛(wèi)星[J];太空探索;2006年10期

相關(guān)會(huì)議論文 前1條

1 王振會(huì);張蕾;楊艷蓉;詹奕哲;張治國(guó);;TCFM導(dǎo)風(fēng)數(shù)據(jù)在臺(tái)風(fēng)分析中的應(yīng)用個(gè)例研究[A];第27屆中國(guó)氣象學(xué)會(huì)年會(huì)雷達(dá)技術(shù)開發(fā)與應(yīng)用分會(huì)場(chǎng)論文集[C];2010年

相關(guān)碩士學(xué)位論文 前2條

1 劉閔;中尺度對(duì)流系統(tǒng)中多個(gè)云團(tuán)的跟蹤和預(yù)測(cè)[D];南京理工大學(xué);2013年

2 朱平;TCFM技術(shù)及其在云導(dǎo)風(fēng)中的應(yīng)用研究[D];南京信息工程大學(xué);2007年

，

本文編號(hào)：1842628