發(fā)布時間：2018-06-15 01:12

  本文選題：MTF + 刃邊法��； 參考：《中國科學(xué)院研究生院（西安光學(xué)精密機(jī)械研究所）》2014年博士論文

【摘要】：光學(xué)遙感成像技術(shù)的飛速發(fā)展，要求空間光學(xué)遙感系統(tǒng)的空間分辨率更高、光譜分辨率更高、輻射分辨率更高，然而遙感器的輕小型化與高分辨率成像之間是矛盾的，這就使得空間光學(xué)遙感器的設(shè)計、研制難度越來越大。受觀測條件的限制，發(fā)送各種條件下的遙感數(shù)據(jù)變得很不現(xiàn)實(shí)，利用實(shí)驗(yàn)室技術(shù)模擬仿真從傳感器、衛(wèi)星平臺、深空環(huán)境、大氣傳輸、地表景物到最終圖像獲取的整個遙感物理過程，能夠保證在經(jīng)濟(jì)成本很低的條件下來獲得客戶所需要的大量圖像。由此可見，光學(xué)遙感仿真技術(shù)在遙感任務(wù)預(yù)測、成像系統(tǒng)設(shè)計、圖像質(zhì)量評估、圖像處理算法驗(yàn)證、圖像解譯訓(xùn)練等一系列領(lǐng)域都具有極其重要的應(yīng)用價值，業(yè)已逐漸成為國內(nèi)外相關(guān)研究機(jī)構(gòu)競相開展的熱門研究課題。 在對光學(xué)遙感系統(tǒng)進(jìn)行建模和仿真之后，我們會得到各種類型的仿真圖像，，需要我們利用各種相應(yīng)的光學(xué)遙感圖像處理技術(shù)對其進(jìn)行處理，從而提取出有用的信息，從而實(shí)現(xiàn)了各種遙感圖像的多樣性應(yīng)用。本文的主要研究工作包括以下幾個方面： 1)基于經(jīng)驗(yàn)總結(jié)的MTF數(shù)學(xué)模型，編寫了基于MTF模型的遙感成像鏈仿真平臺V1.0，實(shí)現(xiàn)在全數(shù)字環(huán)境中動態(tài)模擬大氣、光學(xué)系統(tǒng)、平臺、CCD探測器、離焦、像移以及電子學(xué)系統(tǒng)等各個環(huán)節(jié)對最終遙感像質(zhì)的影響，有助于綜合評價遙感器的成像性能，從而實(shí)現(xiàn)遙感器的優(yōu)化設(shè)計。 2)在基于刃邊法計算遙感圖像的MTF的算法中，提出一種新型的非線性二乘迭代算法擬合Fermi函數(shù)，摒棄了模擬退火算法計算函數(shù)系數(shù)時的缺點(diǎn)，在精度和速度方面都有較大改進(jìn)。 3)基于對各種傳統(tǒng)彩色插值算法的分析和比較，本文提出了一種新型的基于權(quán)重邊緣-權(quán)重色差的彩色插值算法，實(shí)現(xiàn)對Bayer圖像進(jìn)行彩色復(fù)原的過程。該算法將色差法則、加權(quán)定律、自適應(yīng)法則和迭代優(yōu)化算法巧妙地結(jié)合在一起，增加了R、G、B三種顏色分量之間的相關(guān)性，而且插值方向選取合適，有效抑制了彩色摩爾條紋。 4)對天文星圖進(jìn)行預(yù)處理、星點(diǎn)粗定位、星點(diǎn)細(xì)定位之后，本文提出了一種新的特征匹配法以全等三角形的重心為特征點(diǎn)。然后基于這些三角形重心點(diǎn)集合，利用RANSAC算法計算配準(zhǔn)轉(zhuǎn)換模型，完成配準(zhǔn)過程。整個過程是利用GPU對星圖配準(zhǔn)算法進(jìn)行了并行程序設(shè)計，并利用Matlab中的nvmex編譯GPU代碼，使得Matlab可以直接調(diào)用生成的MEX文件，有效的加速了天文星圖的配準(zhǔn)速度，總時間加速比達(dá)到29.043倍。 5)利用自制的多波段偏振CCD地面實(shí)驗(yàn)裝置，將白熾燈作為照明光源，獲取目標(biāo)的偏振光譜圖像。本文實(shí)驗(yàn)的目的是辨別真花和假花、真葉和假葉，以DOLP為評價指標(biāo)，發(fā)現(xiàn)與灰度圖像相比，偏振度圖像中的真花和假花的對比度數(shù)值更大，即兩者更容易被區(qū)分。
[Abstract]:The rapid development of optical remote sensing imaging technology requires higher spatial resolution, higher spectral resolution and higher radiation resolution of space optical remote sensing system. However, there is a contradiction between the miniaturization of remote sensor and high resolution imaging. This makes the design and development of space optical remote sensor more and more difficult. Limited by observation conditions, it is very unrealistic to send remote sensing data under various conditions, using laboratory technology to simulate and simulate the transmission from sensors, satellite platforms, deep space environments, atmosphere, The whole physical process of remote sensing from the surface to the final image can ensure that the large number of images needed by the customer can be obtained under the condition of low economic cost. It can be seen that optical remote sensing simulation technology has very important application value in a series of fields such as remote sensing task prediction, imaging system design, image quality evaluation, image processing algorithm verification, image interpretation training and so on. It has gradually become a hot research topic that related research institutions at home and abroad are competing to carry out. After modeling and simulating the optical remote sensing system, we will get various kinds of simulation images, which need to be processed by various optical remote sensing image processing techniques, so as to extract useful information. Thus, the diversity of remote sensing images is realized. The main research work of this paper includes the following aspects: 1) based on the MTF mathematical model, a remote sensing imaging chain simulation platform V1.0 based on MTF model is developed to realize the dynamic simulation of atmosphere and optical system in the full digital environment. The effects of various aspects of the platform CCD detector, defocus, image shift and electronic system on the final remote sensing image quality are helpful for the comprehensive evaluation of the imaging performance of the remote sensor. 2) in the algorithm of calculating MTF of remote sensing image based on edge method, a new nonlinear two-multiplication iterative algorithm is proposed to fit Fermi function. In this paper, the shortcomings of simulated annealing algorithm in calculating function coefficients are abandoned, and the accuracy and speed are improved greatly. 3) based on the analysis and comparison of various traditional color interpolation algorithms, In this paper, a new color interpolation algorithm based on weighted edge and weight color difference is proposed to realize the color restoration of Bayer image. The algorithm combines color difference law, weighting law, adaptive rule and iterative optimization algorithm skillfully, which increases the correlation among the three color components, and the direction of interpolation is suitable. In this paper, we propose a new feature matching method, which is based on the center of gravity of a full triangle, after the pretreatment of star map, the coarse location of star and the fine location of star. Then, based on the set of triangle barycenter points, the registration transformation model is calculated by using RANSAC algorithm, and the registration process is completed. The whole process is to make use of GPU to design the parallel program of star map registration algorithm, and compile GPU code by using nvmex in Matlab, so that Matlab can directly call the generated MEX file, which effectively accelerates the registration speed of astronomical star map. The acceleration ratio of total time is 29.043 times. 5) using the self-made multi-band polarization CCD ground experimental device, the incandescent lamp is used as the illuminating light source to obtain the polarizing spectrum image of the target. The purpose of this experiment is to distinguish true flower from false flower, true leaf and false leaf. Using DOLP as the evaluation index, we find that the contrast value of true flower and false flower in polarization degree image is larger than that in gray image, that is to say, it is easier to distinguish the true flower from false flower.
【學(xué)位授予單位】：中國科學(xué)院研究生院（西安光學(xué)精密機(jī)械研究所）
【學(xué)位級別】：博士
【學(xué)位授予年份】：2014
【分類號】：TP751

【參考文獻(xiàn)】

中國期刊全文數(shù)據(jù)庫 前10條

1 張冬英;易維寧;王先華;喬延利;;空基遙感圖像的MTF試驗(yàn)研究[J];安徽大學(xué)學(xué)報(自然科學(xué)版);2007年06期

2 羅志灶;周贏武;鄭忠楷;;二值圖像連通域標(biāo)記優(yōu)化算法[J];安慶師范學(xué)院學(xué)報(自然科學(xué)版);2010年04期

3 范鵬飛;范秀英;梁勇;楊必軍;;基于成像過程的航空相機(jī)成像仿真系統(tǒng)[J];兵工自動化;2010年11期

4 魏新國,張廣軍,江潔;星敏感器中星圖圖像的星體細(xì)分定位方法研究[J];北京航空航天大學(xué)學(xué)報;2003年09期

5 林躍春;郭金海;雷陽;;淺析偏振光遙感的應(yīng)用[J];測繪與空間地理信息;2011年02期

6 賈志科,崔慧娟,唐昆;彩色數(shù)字照相機(jī)的設(shè)計[J];電視技術(shù);2000年11期

7 宋斌;;一種新的圖像連通域快速標(biāo)號算法[J];電子測量技術(shù);2009年09期

8 蔣剛毅;黃大江;王旭;郁梅;;圖像質(zhì)量評價方法研究進(jìn)展[J];電子與信息學(xué)報;2010年01期

9 蔣龍聰;劉江平;;模擬退火算法及其改進(jìn)[J];工程地球物理學(xué)報;2007年02期

10 王鉦旋,李志林,龐云階;一個二值圖像連通成份標(biāo)記的快速算法[J];工程圖學(xué)學(xué)報;1998年03期

中國博士學(xué)位論文全文數(shù)據(jù)庫 前6條

1 藍(lán)朝楨;空間目標(biāo)天基光學(xué)觀測系統(tǒng)建模與探測能力分析[D];解放軍信息工程大學(xué);2009年

2 肖漢;基于CPU+GPU的影像匹配高效能異構(gòu)并行計算研究[D];武漢大學(xué);2011年

3 杜西亮;基于振幅分割的光偏振測量技術(shù)的研究[D];哈爾濱工業(yè)大學(xué);2007年

4 單列;視覺注意機(jī)制的若干關(guān)鍵技術(shù)及應(yīng)用研究[D];中國科學(xué)技術(shù)大學(xué);2008年

5 付強(qiáng);基于成像鏈分析的光譜成像系統(tǒng)設(shè)計方法研究[D];中國科學(xué)院研究生院(西安光學(xué)精密機(jī)械研究所);2012年

6 黃濤;基于GPU的多點(diǎn)地質(zhì)統(tǒng)計逐點(diǎn)模擬并行算法的研究[D];中國科學(xué)技術(shù)大學(xué);2013年

本文編號：2019854