本文選題：DBN + SAR��； 參考：《西安電子科技大學(xué)》2015年碩士論文

【摘要】：深度學(xué)習(xí)起源于人工神經(jīng)網(wǎng)絡(luò),模仿人腦計(jì)算模式,可以自動(dòng)地分層學(xué)習(xí)出抽象特征,在圖像領(lǐng)域應(yīng)用廣泛,尤其是在目標(biāo)識(shí)別和圖像分類方面。隨著遙感技術(shù)的發(fā)展,合成孔徑雷達(dá)(Synthetic aperture radar,SAR)圖像以其大信息量,全天候全天時(shí)的特點(diǎn)在軍事和民用領(lǐng)域占據(jù)了重要地位。對(duì)于SAR圖像的識(shí)別和分類任務(wù)而言,選取合適的特征非常重要,所用特征決定了算法性能的上限,而深度學(xué)習(xí)模型可以自動(dòng)從原始數(shù)據(jù)中學(xué)出更抽象的特征。在深度學(xué)習(xí)模型中,深度置信網(wǎng)絡(luò)(Deep Belief Network,DBN)使用的棧式限制玻爾茲曼機(jī)(Restricted Boltzmann Machine,RBM)的非監(jiān)督學(xué)習(xí)和反向傳播的有監(jiān)督微調(diào)過(guò)程,可以自動(dòng)學(xué)習(xí)到更適合分類的特征。本文用深度置信網(wǎng)絡(luò)來(lái)提取高層抽象特征用于SAR圖像。具體工作如下:一,用于單極化SAR目標(biāo)識(shí)別的深度學(xué)習(xí)研究。由于深度學(xué)習(xí)更適用于大數(shù)據(jù),而本文所用的運(yùn)動(dòng)與靜止目標(biāo)的獲取與識(shí)別數(shù)據(jù)(MSTAR)數(shù)量有限,使得深度學(xué)習(xí)模型不容易收斂,所以我們提出數(shù)據(jù)融合與深度學(xué)習(xí)相結(jié)合的策略。分別提取MSTAR的輪廓波特征和曲線波特征與原數(shù)據(jù)相結(jié)合作為深度置信網(wǎng)絡(luò)的輸入,同時(shí)加入更能模擬數(shù)據(jù)的高斯限制玻爾茲曼機(jī)(gaussianRBM),進(jìn)行SAR圖像的目標(biāo)識(shí)別,識(shí)別精度較原始RBM和單一數(shù)據(jù)有所提高。由于傳統(tǒng)DBN沒(méi)有考慮到SAR圖像的2-D結(jié)構(gòu)和空間信息,導(dǎo)致學(xué)習(xí)到的權(quán)值與像素所處位置無(wú)關(guān),而卷積網(wǎng)絡(luò)的權(quán)值共享使得每一種權(quán)值對(duì)應(yīng)一種特征算子,更利于提取不同性質(zhì)的特征,所以,本文使用基于卷積RBM的深度置信網(wǎng)絡(luò),使得識(shí)別精度進(jìn)一步提高。二,用于全極化SAR圖像地物分類的深度學(xué)習(xí)研究。由于傳統(tǒng)的RBM更適合模擬二值數(shù)據(jù),對(duì)于符合其他指數(shù)家族的分布,RBM可以加入不同的統(tǒng)計(jì)特性進(jìn)行擴(kuò)展。所以,對(duì)極化SAR實(shí)數(shù)數(shù)據(jù),我們使用加入高斯分布的gaussianRBM構(gòu)成DBN,用于極化SAR圖像的地物分類;對(duì)極化SAR復(fù)數(shù)數(shù)據(jù),我們基于極化SAR數(shù)據(jù)復(fù)wishart分布特性提出了的wishartRBM,并由此構(gòu)成DBN,用于極化SAR圖像的地物分類。具體步驟為:將極化SAR數(shù)據(jù)的協(xié)方差矩陣元素作為輸入,先用多層的wishartRBM預(yù)訓(xùn)練網(wǎng)絡(luò),再加上反向傳播進(jìn)行微調(diào),最后使用softmax分類進(jìn)行地物分類,分類精度與其他方法相比得到了提高。
[Abstract]:Depth learning originates from artificial neural network and imitates the human brain computing model. It can automatically learn abstract features in layers and is widely used in image field, especially in target recognition and image classification. With the development of remote sensing technology, synthetic aperture radar synthetic aperture radar (SAR) images play an important role in military and civilian fields with its large amount of information and all-weather and all-day characteristics. For the task of SAR image recognition and classification, it is very important to select suitable features, which determine the upper bound of the algorithm performance, while the depth learning model can automatically learn more abstract features from the original data. In the deep learning model, the unsupervised learning and backpropagation process of unsupervised learning and backpropagation of the deep confidence network Deep Belief Network (DBN) can automatically learn more suitable features for classification. In this paper, a depth confidence network is used to extract high-level abstract features for SAR images. The main work is as follows: 1. The research of deep learning for single polarization SAR target recognition. Because depth learning is more suitable for big data, and the number of moving and static target acquisition and recognition data is limited in this paper, it is difficult to converge in depth learning model, so we propose a combination of data fusion and depth learning strategy. The contour wave feature, curve wave feature and original data of MSTAR are extracted respectively as input of depth confidence network, and Gao Si restricted Boltzmann machine, which can simulate data, is added to realize target recognition of SAR image. The recognition accuracy is improved compared with the original RBM and single data. Because the traditional DBN does not consider the 2-D structure and spatial information of the SAR image, the weights learned are independent of the location of the pixels, and the weights of the convolutional network share the weights so that each weight corresponds to a feature operator. Therefore, the depth confidence network based on convolution RBM is used to improve the recognition accuracy. Secondly, it is used to study the depth learning of ground object classification in fully polarized SAR images. Because the traditional RBM is more suitable for simulating binary data, different statistical properties can be added to the distributed RBM which accords with other exponential families. So, for the real data of polarimetric SAR, we use the gaussianRBM with Gao Si distribution to form DBNs, which is used to classify the ground objects in polarized SAR images, and for the complex data of polarimetric SAR, Based on the complex wishart distribution of polarized SAR data, we propose wishart RBM, which is used to classify ground objects in polarimetric SAR images. The concrete steps are as follows: the element of covariance matrix of polarized SAR data is taken as input, the multi-layer wishartRBM pre-training network is used first, then the backpropagation is used to fine tune, and finally, softmax classification is used to classify ground objects. The classification accuracy is improved compared with other methods.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：TN957.52

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 沙宇恒,叢琳,孫強(qiáng),焦李成;基于Contourlet域HMT模型的多尺度圖像分割[J];紅外與毫米波學(xué)報(bào);2005年06期

2 焦李成,譚山;圖像的多尺度幾何分析:回顧和展望[J];電子學(xué)報(bào);2003年S1期

，

本文編號(hào)：1851157