本文選題：高光譜圖像 + 卷積神經(jīng)網(wǎng)絡(luò)　； 參考：《電子科技大學(xué)》2017年碩士論文

【摘要】：遙感技術(shù)的發(fā)展使得新的遙感傳感器可以采集的圖像具有連續(xù)的譜域和空域,這些圖像含有大量的地物信息——地物的光譜信息和幾何空間分布。傳統(tǒng)的遙感圖像分類框架只利用了譜域信息來分類,忽略了空域信息對分類的影響。卷積神經(jīng)網(wǎng)絡(luò)有著獨(dú)特的優(yōu)勢。圖像不需要太多的前期處理就可以直接輸入網(wǎng)絡(luò),從訓(xùn)練數(shù)據(jù)中隱式地進(jìn)行學(xué)習(xí),規(guī)避了特征提取和分類中數(shù)據(jù)重建的過程。它獨(dú)特的層間聯(lián)系以及空間信息的密切聯(lián)系,使其適用于圖像處理中分類識別任務(wù)。本文在充分考慮高光譜圖像特點(diǎn)的前提下,提出了在圖像分類領(lǐng)域中取得顯著成果的卷積神經(jīng)網(wǎng)絡(luò)來對高光譜圖像的像元進(jìn)行分類。本文主要研究內(nèi)容:(1)借鑒LeNet-5網(wǎng)絡(luò)框架的設(shè)計(jì)思想,設(shè)計(jì)適用于高光譜圖像分類的卷積神經(jīng)網(wǎng)絡(luò)框架。本文主要研究了網(wǎng)絡(luò)框架中的層數(shù)、卷積層中神經(jīng)元的數(shù)目、下采樣層中神經(jīng)元的數(shù)目以及輸出層中神經(jīng)元的數(shù)目,達(dá)到了高光譜數(shù)據(jù)分類到圖像分類的有效轉(zhuǎn)化。(2)將每個像素點(diǎn)的空間鄰域信息作為卷積神經(jīng)網(wǎng)絡(luò)框架的輸入樣本,探究所設(shè)計(jì)的框架對高光譜圖像分類的有效性。(3)研究框架中的激活函數(shù)ReLU的設(shè)計(jì),達(dá)到緩解梯度彌散目的,提高網(wǎng)絡(luò)的執(zhí)行效率和分類精度。與梯度下降法相比,mini-batch隨機(jī)梯度下降法可以大大提高框架的執(zhí)行效率。這些策略的使用更有助于提取分類特征和提高分類效果。(4)利用本文所設(shè)計(jì)的框架對The University of Pavia數(shù)據(jù)集進(jìn)行分類仿真實(shí)驗(yàn),驗(yàn)證其可行性,并與傳統(tǒng)的k近鄰、BP神經(jīng)網(wǎng)絡(luò)以及SVM分類識別方法比較。實(shí)驗(yàn)仿真結(jié)果表明:采用本文所設(shè)計(jì)的分類框架,分類精度高于其它分類方法,達(dá)到了97.57%。
[Abstract]:With the development of remote sensing technology, the images which can be collected by the new remote sensing sensor have continuous spectral domain and spatial domain. These images contain a lot of ground object information-spectral information and geometric spatial distribution. The traditional remote sensing image classification framework only uses spectral domain information to classify, ignoring the influence of spatial information on classification. Convolutional neural networks have unique advantages. The image can be directly input into the network without too much pre-processing, and can be learned implicitly from the training data, thus avoiding the process of feature extraction and data reconstruction in classification. It is suitable for classification and recognition in image processing because of its unique interlayer connection and close connection of spatial information. On the premise of fully considering the characteristics of hyperspectral images, a convolution neural network, which has achieved remarkable results in the field of image classification, is proposed in this paper to classify the pixels of hyperspectral images. In this paper, we use the design idea of LeNet-5 network framework for reference, and design a convolution neural network framework suitable for hyperspectral image classification. In this paper, the number of layers in the network framework, the number of neurons in the convolution layer, the number of neurons in the lower sampling layer and the number of neurons in the output layer are studied. An effective transformation from hyperspectral data classification to image classification is achieved. The spatial neighborhood information of each pixel is used as the input sample of the convolution neural network framework. The effectiveness of the proposed framework for hyperspectral image classification is explored. (3) the design of the activation function (ReLU) in the framework is studied in order to alleviate the gradient dispersion and improve the efficiency and classification accuracy of the network. Compared with the gradient descent method, the mini-batch stochastic gradient descent method can greatly improve the performance efficiency of the frame. The use of these strategies is more helpful to extract the classification features and improve the classification effect. (4) using the framework designed in this paper, the The University of Pavia dataset classification simulation experiments are carried out to verify its feasibility. And compared with the traditional k-nearest neighbor BP neural network and SVM classification and recognition methods. The experimental results show that the classification accuracy of the proposed classification framework is higher than that of other classification methods, reaching 97.57.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP751

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 關(guān)海鷗;杜松懷;許少華;左豫虎;;基于改進(jìn)投影尋蹤技術(shù)和模糊神經(jīng)網(wǎng)絡(luò)的未受精種蛋檢測模型[J];江蘇大學(xué)學(xué)報(bào)(自然科學(xué)版);2013年02期

2 李海峰;李純果;;深度學(xué)習(xí)結(jié)構(gòu)和算法比較分析[J];河北大學(xué)學(xué)報(bào)(自然科學(xué)版);2012年05期

3 王國林;周樹仁;李軍強(qiáng);;基于模糊聚類和形態(tài)學(xué)的輪胎斷面特征提取[J];江蘇大學(xué)學(xué)報(bào)(自然科學(xué)版);2012年05期

4 王立國;孫杰;肖倩;;結(jié)合空-譜信息的高光譜圖像分類方法[J];黑龍江大學(xué)自然科學(xué)學(xué)報(bào);2010年06期

5 張睿;馬建文;;支持向量機(jī)在遙感數(shù)據(jù)分類中的應(yīng)用新進(jìn)展[J];地球科學(xué)進(jìn)展;2009年05期

6 譚琨;杜培軍;;基于支持向量機(jī)的高光譜遙感圖像分類[J];紅外與毫米波學(xué)報(bào);2008年02期

7 田野;趙春暉;季亞新;;主成分分析在高光譜遙感圖像降維中的應(yīng)用[J];哈爾濱師范大學(xué)自然科學(xué)學(xué)報(bào);2007年05期

8 儲美華;洪志剛;陳允芳;;數(shù)字航測影像盲點(diǎn)消除與病態(tài)像元校正方法探討[J];測繪科學(xué);2007年03期

9 王立國,張曄,谷延鋒;支持向量機(jī)多類目標(biāo)分類器的結(jié)構(gòu)簡化研究[J];中國圖象圖形學(xué)報(bào);2005年05期

10 呂東亞,黃普明,孫獻(xiàn)璞;高光譜圖像的數(shù)據(jù)特征及壓縮技術(shù)[J];空間電子技術(shù);2005年01期

相關(guān)博士學(xué)位論文 前1條

1 李向軍;遙感土地利用變化檢測方法探討[D];中國科學(xué)院研究生院（遙感應(yīng)用研究所）;2006年

相關(guān)碩士學(xué)位論文 前1條

1 孫麗娟;基于支持向量機(jī)的高光譜圖像分類技術(shù)研究[D];哈爾濱工程大學(xué);2011年

，

本文編號：1775499