遙感專業(yè)外文翻譯--高光譜遙感信息中的特征提取與應(yīng)用研究

1、<p>　　2500單詞，3900漢字</p><p>　　出處：Du P, Tao F, Hong T. Spectral Features Extraction in Hyperspectral RS Data and Its Application to Information Processing [J]. Acta Photonica Sinica, 2005, 34(2):293-298.

2、</p><p>　　本科畢業(yè)設(shè)計(jì)（論文）</p><p><b>　　中英文對(duì)照翻譯</b></p><p>　　院（系部） 測(cè)繪與國(guó)土信息工程學(xué)院 </p><p>　　專業(yè)名稱 測(cè)繪工程 　 </p><p>　　年級(jí)班級(jí)　

3、</p><p>　　學(xué)生姓名　 </p><p>　　指導(dǎo)老師 </p><p><b>　　2012年6月3日</b></p><p>　　Spectral Features Extraction in Hyperspectr

4、al RS Data and</p><p>　　Its Application to Information Processing</p><p>　　Oriented to the demands of hyperspectral RS information processing and applications, spectral features in hyperspectral

5、 RS image can be categorized into three scales: point scale, block scale and volume scale. Based on the properties and algorithms of different features, it is proposed that point scale features can be divided into three

6、levels: spectral curve features, spectral transformation features and spectral similarity measure features. Spectral curve features include direct spectra encodi</p><p>　　1 Introduction</p><p>　

7、　Hyperspectral Remote Sensing was one of the most important breakthroughs of Earth Observation System ( EOS) in 1990 s. It overcomes the limitations of conventional aerial and multispectral RS such as less band amount, w

8、ide band scope and rough spectral information expression, and can provide RS information with narrow band width, more band amount and fine spectral information, also it can distinguish and identify ground objects from sp

9、ectral space, so hyperspectral RS has got wide applications i</p><p>　　2 Framework of spectral features in hyperspectral RS data</p><p>　　In general, hyperspectral RS image can be expressed by a

10、 spatial-spectral data cube ( Fig. 1). In this data cube, every coverage expressed the image of one band, and each pixel forms a spectral vector composed of albedo of ground object on every band in spectral dimension, an

11、d that vector can be visualized by spectral curve ( Fig. 2 ). Many features can be extracted from spectral vector or curve, and spectral features are the key and basis of hyperspectral RS applications. Also each spectral

12、 cur</p><p>　　Fig. 1　Hyperspectral image data cube Fig. 2　Reflectance spectral curve of a pixel</p><p>　　2. 1　Three scales of spectral features</p><p>　　According to the ope

13、rational objects of extraction algorithms, spectral features can be categorized into three scales: point-scale, block-scale and volume-</p><p><b>　　Scale.</b></p><p>　　Point scale ta

14、kes pixel and its spectral curve as operational object and some useful features can be extracted from this spectral vector (or spectral curve).In general, hyperspectral RS image takes spectral vector of each pixel as pro

15、cessing object.</p><p>　　Block scale is oriented image block or region. Block is the set of some pixels, and it can be homogeneous or heterogeneous. Homogeneous regions are got by image segmentation and pixe

16、ls in this region are similar in some given features; heterogeneous region are those image blocks with regular or irregular size, and they are cut from original image directly, for example, an image can be segmented acco

17、rding to quadtree method. In hyperspectral RS image, block scale features can be computed from two</p><p>　　Volume scale combines spatial and spectral features in a whole and extracts features in 3D ( row, c

18、olumn and spectra ) space. Here, some 3D operational algorithms are needed, for example, 3D wavelet transformation and high order Artificial Neural Network (ANN ). Because this type of features is difficult to compute an

19、d analyze, we don′t research it in current studies.</p><p>　　In this paper, we would like to focus on point scale feature, or those features extracted from spectral vector that may be spectral vector of a pi

20、xel or mean vector of a block.</p><p>　　2. 2　Three levels of point scale features</p><p>　　From operation object, algorithm principles, feature properties, application modes and other aspects, w

21、e think it is feasible to categorize spectral features into three levels: spectral curve features, spectral transformation features and spectral similarity measure features. They are corresponding to analysis on spectral

22、 curve with all bands, data transformation and combination with part of all bands and similarity measure of spectral vectors. In our study, data from OM IS and PHI hyperspectral</p><p>　　3　Spectra l curve fe

23、atures</p><p>　　Spectral curve features are computed by some algorithms based on the spectral curve of certain pixel or ground object, and it can describe shape and properties of the curve. The main methods

24、include direct encoding and feature band analysis.</p><p>　　3. 1　Direct encoding</p><p>　　The important idea of spectral curve feature is to emphasize spectral curve shape, so direct encoding is

25、 a very convenient method, and binary encoding is used more widely. Its principle is to compare the attribute value at each band of a pixel with a threshold and assign the code of“0”or“1”according to its value. That can

26、be expressed by</p><p>　　Here, is code of the ith band, is the original attribute value of this band, and T is the threshold. Generally, threshold is the mean of spectral vector, and it can also be selected

27、 by manual method according to curve shape, sometimes median of spectral vector is probably used.</p><p>　　Only one threshold is used in binary encoding, so the divided internal is large and precision is low

28、. In order to improve the appoximaty and precision, the quaternary encoding strategy is proposed in this paper. Its primary idea is as follows: ( 1 ) the mean of the total pixel spectral vector is computed and denoted by

29、 T0 , and the attribute is divided into two internal including [ , ] and [ , ]; (2) the pixels located in the two internals are determined and the mean of each internal is got an</p><p>　　If quarternary en

30、coding is used, the ratio of the same region is smaller than binary encoding, but the ratio between different regions decreased dramatically. So quarternary encoding is more effective in measuring the similarity between

31、different pixels. </p><p>　　Because direct encoding will disperse the continuous albedo into discrete code, the encoding result is affected by threshold obviously and will lead to information loss. Although

32、its operation is very simple, it is only used to some applications requiring low precision, and the threshold should be selected according to different conditions.</p><p>　　3. 2　Spectral absorption or reflec

33、tion feature</p><p>　　Differing from direct encoding in which all bands are used, spectral absorption or reflection feature only emphasizes those bands where valleys or apexes are located. That means those b

34、ands with local maximum or minimum in spectral curve should be determined at first and then further analysis can be done. In general, albedo is used to describe the attribute of a pixel, so those bands with local maximum

35、 are reflection apex and those with local minimum are absorption valley.</p><p>　　After the location and related parameters are got, the detail analysis can be done. In general two methods are used, one is t

36、o give direct encoding and analysis to feature bands, and the other is to compute some quantitative index using feature bands and their parameters.</p><p>　　3.3　Encoding of spectra l absorption or reflection

37、 features</p><p>　　The locations of feature bands are directly used in spectral feature encoding. The following will take absorption feature as an example. If one band is the location of absorption valley, i

38、ts code will be “1 ”, otherwise its code is “0 ”. After the encoding is completed further matching and comparison can be done. Because of those uncertainties and errors in hyper spectral imaging process, the locations of

39、 feature bands perhaps move in near bands, and that will lead to low match ratio. In order t</p><p>　　The similarity measure to code vector is matching by bit. The matching ratio is got by the ratio of match

40、ed bands to total band count. In this study, two match schemes are used. One is matching the code of all bands and the other is only matching those feature bands.</p><p>　　Based on above analysis, four schem

41、es are used and compared. These are: ( 1) direct encoding to all bands and matching by all bands, and ( 2 ) direct encoding to all bands and matching only by feature bands, and ( 3) extended encoding and matching by all

42、bands, and ( 4 ) extended encoding and matching only by feature bands. </p><p>　　From above analysis and comparison to spectral absorption and reflection feature encoding and matching, it can be found that

43、although absorption and reflection band can describe the spectral properties of ground object, effective matching operation should be used in order to overcome the impacts of noise, band displacement and other factors. I

44、n practical applications, absorption and reflection can be used to extract thematic information and retrieve a certain type of object effectively.</p><p>　　Based on spectral absorption and reflection feature

45、s, the spectral absorption index ( SA I) or spectral reflection index ( SR I) can be computed by wavelength, albedo of feature band and its left and right shoulders, and those indexes can describe spectral feature more

46、 precisely on some occasions.</p><p>　　4 Spectra l computation and transformation features</p><p>　　Both correlativity and mutual compensation exist in different bands of hyper spectral RS infor

47、mation, so many new features can be got by certain computation and combination to some bands and used to classification, information extraction and other tasks.</p><p>　　4. 1 　Normalized difference of vegeta

48、tion index (NDVI)</p><p>　　NDVI plays very important roles in hyper spectral application. It can describe some fine information about vegetation such as Leaf Area Index (LA I) , ratio of vegetation and soil,

49、 component of vegetation and so on. In some classifiers ( for example, ANN classifier) NDVI usually is used as an independent feature in classification.</p><p>　　4. 2　Derivative spectrum</p><p>

50、;　　Derivative spectrum is also called as spectral derivative technique. One rank and two rank derivative spectrum can be computed by Equation.</p><p>　　Each rank derivative spectrum can be computed using alg

51、orithms similar to above. After derivative computation is end, we can find that each type of ground object may have some features distinguished from other entities in a certain rank derivative spectrum and that can be us

52、ed to identify information. Sometimes derivative spectrum image can be used as the input of classifier directly. Although spectral derivative can provide new features in addition to original information, some new images

53、will </p><p>　　5　Conclusions and discussions</p><p>　　In this paper, oriented to the demands of hyper spectral RS information processing to spectral features, the framework of spectral features

54、is proposed and some major feature extraction algorithms and their applications are discussed, and some improvement, experiments and analysis are finished. From the studies in this paper, the following conclusions can be

55、 drawn:</p><p>　　1 ) Based on the extraction principle and algorithm, spectral features in hyper spectral RS information can be categorized into three levels: spectral curve features, spectral transformation

56、 and computation features and spectral similarity measure features. This framework is useful for further analysis and applications.</p><p>　　2) As the common style of pixel spectral vector, some features can

57、 be extracted and used. The algorithm and computation of binary encoding is simple and easy but it will lead to loss of some detail information. Quaternary encoding can describe curve features with high rescission and b

58、e used to matching, retrieval and other work. The reflection and absorption features based on spectral curve have wide applications in retrieval, thematic information extraction and other tasks, but effective match</p

59、><p>　　3) As the main computation and transformation features, NDV I and derivative spectrum can provide new features participating in classification, extraction and other processing and extract those useful pa

60、tterns and information hidden behind original data, so they are very useful in hyper spectral RS information processing.</p><p>　　4) For those spectra similarity measure indexes, Spectral Angle and SID are m

61、ore effective than traditional indexes because they can measure the similarity more precisely, so they are usually used to classification, clustering and retrieval.</p><p>　　Some topics about the feature ext

62、raction and application of spectral feature are discussed in this paper. Our further studies will be focused on classification, object identification and thematic information extraction in hyper spectral RS information a

63、nd the specific application modes of different spectral features in order to promote the development of hyper spectral RS application.</p><p>　　高光譜遙感信息中的特征提取與應(yīng)用研究</p><p>　　面向高光譜遙感信息處理和應(yīng)用的需求，在高光譜

64、遙感圖像的光譜特征可分為三個(gè)尺度：點(diǎn)規(guī)模，塊規(guī)模和數(shù)量規(guī)模。根據(jù)不同的功能屬性和算法，它提出了點(diǎn)規(guī)模的特點(diǎn)，可以分為三個(gè)層次：光譜曲線特征，光譜變換特征和光譜相似性度量功能。光譜曲線特征包括直接光譜編碼，反射和吸收功能。光譜變換特征包括歸一化植被指數(shù)（紐卡斯?fàn)枺瑢?dǎo)數(shù)光譜和其他光譜的計(jì)算功能。光譜相似性度量的功能包括光譜角（SA），光譜信息散度（SID）的光譜距離，相關(guān)系數(shù)等。分析這些算法的幾個(gè)問題，關(guān)于特征提取，匹配和應(yīng)用的基礎(chǔ)上進(jìn)一

65、步討論，它分析得到的第四紀(jì)編碼，光譜角和SID可有效用于信息處理。</p><p><b>　　1介紹</b></p><p>　　高光譜遙感是在1990年的地球觀測(cè)系統(tǒng)（EOS），最重要的突破之一。它克服了傳統(tǒng)的天線和多光譜RS，如少帶量，寬波段范圍和粗糙的光譜信息表達(dá)的限制，可以提供寬窄帶，更帶量和良好的光譜信息的遙感信息，還可以區(qū)分和識(shí)別地面光譜空間中的對(duì)象，因

66、此高光譜遙感在資源，環(huán)境，城市和生態(tài)領(lǐng)域得到廣泛應(yīng)用。由于高光譜遙感是從傳統(tǒng)的遙感信息的信息采集和信息處理明顯不同，有許多問題要在實(shí)踐中加以解決。最重要的問題之一是高光譜遙感圖像與標(biāo)準(zhǔn)光譜數(shù)據(jù)庫(kù)的光譜遙感數(shù)據(jù)光譜特征提取和應(yīng)用。如今，在光譜的研究主要集中在波段選擇和組合，圖像分類，混合像元分解和他人，和光譜特征的研究很少。在本文中，光譜特征的提取和應(yīng)用將作為我們的中心議題，以高光譜遙感應(yīng)用提供一些有益的建議。</p>&l

67、t;p>　　2高光譜遙感數(shù)據(jù)光譜特征的框架</p><p>　　一般情況下，高光譜遙感圖像可以表示空間光譜數(shù)據(jù)立方體（圖1）。覆蓋在這個(gè)多維數(shù)據(jù)集，每一個(gè)面的形象，每個(gè)像素形成一個(gè)光譜，每個(gè)波段的光譜維矢量組成的地面物體的反照率，可以通過光譜曲線（圖2）可視化，矢量化。許多功能可以提取光譜向量或曲線，光譜特征的高光譜遙感應(yīng)用的關(guān)鍵和基礎(chǔ)。每個(gè)光譜數(shù)據(jù)庫(kù)的光譜曲線也可以用同樣的方法分析。雖然有一些算法來(lái)計(jì)算的

68、光譜特征，框架和體系還不夠明顯，所以我們想提出一個(gè)包括高光譜遙感圖像與標(biāo)準(zhǔn)光譜數(shù)據(jù)庫(kù)的光譜遙感數(shù)據(jù)光譜特征的框架。</p><p>　　圖1高光譜圖像數(shù)據(jù)立方體 圖2像素的反射光譜曲線</p><p>　　2.1三個(gè)尺度的光譜特征</p><p>　　據(jù)光譜特征提取算法的運(yùn)作對(duì)象，可以分為三個(gè)尺度：點(diǎn)規(guī)模，塊規(guī)模和數(shù)量規(guī)模。點(diǎn)規(guī)模的

69、像素和其光譜曲線的經(jīng)營(yíng)對(duì)象及一些有用的功能，可以從這個(gè)光譜的矢量（或光譜曲線）中提取。一般情況下，高光譜遙感圖像每個(gè)像素的光譜向量作為處理對(duì)象。</p><p>　　塊規(guī)模為導(dǎo)向的圖像塊或地區(qū)。塊是一些像素的集合，它可以是同質(zhì)或異質(zhì)。同質(zhì)區(qū)域的圖像分割，并在本地區(qū)的像素了，在一些特定的功能相似；異構(gòu)地區(qū)定期或不定期的大小的圖像塊，他們從原來(lái)的圖像直接削減，例如，圖像可以被分割根據(jù)四叉樹法。在高光譜遙感圖像，塊尺度

70、特征可以從兩個(gè)方面來(lái)計(jì)算：一個(gè)是計(jì)算塊的紋理特征，在某些特點(diǎn)的與運(yùn)用，另一種是計(jì)算塊的光譜特征，如果該塊是均勻的，首先可以計(jì)算其均值向量，然后可以提取光譜，這叫矢量描述塊。如果塊是異構(gòu)的，它可以分割成一些均勻的塊。</p><p>　　批量規(guī)模結(jié)合在一個(gè)整體包括光譜特征和提取3D功能（行，列和光譜）的空間。在這里，一些3D的操作算法是必要的，例如，三維小波變換和高階人工神經(jīng)網(wǎng)絡(luò)（ANN）。因?yàn)檫@種類型的特點(diǎn)是很難

71、計(jì)算和分析，我們不研究在當(dāng)前的研究。</p><p>　　本文中，我們希望把重點(diǎn)放在功能點(diǎn)規(guī)模，或從光譜向量，因?yàn)橐粋€(gè)像素的光譜向量或向量意味著一個(gè)塊中提取這些功能。</p><p>　　2.2三個(gè)層次點(diǎn)尺度特征</p><p>　　從操作對(duì)象，算法原理，功能特性，應(yīng)用模式等方面，我們認(rèn)為這是可分為三個(gè)層次光譜特性：光譜曲線特征，光譜變換特征和光譜相似性度量功能。他

72、們是對(duì)應(yīng)的所有頻段，數(shù)據(jù)轉(zhuǎn)換，并與所有波段和光譜向量相似性度量的一部分相結(jié)合的光譜曲線分析。在我們的研究中，從奧姆數(shù)據(jù)的PHI高光譜圖像看出，美國(guó)地質(zhì)勘探局的光譜數(shù)據(jù)庫(kù)和典型的光譜數(shù)據(jù)，在中國(guó)的試驗(yàn)和本文給出了兩個(gè)例子。一個(gè)是選擇三個(gè)地區(qū)，從PH我的形象（我是植被的地區(qū)，二區(qū)是建筑物的土地，和地區(qū)三是一些土地面積混合區(qū)），另一種是從美國(guó)地質(zhì)勘探局的光譜數(shù)據(jù)庫(kù)，其中三個(gè)地面物體的光譜曲線中S1是陽(yáng)起石的HS22。 3B，S2是陽(yáng)起石的HS

73、116。3B和S3是陽(yáng)起石的HS66。 3B，S1和S2是相似的，與S3不同。</p><p><b>　　3光譜曲線功能</b></p><p>　　計(jì)算光譜曲線特征的基礎(chǔ)上，某些像素或地面物體的光譜曲線的一些算法，它可以描述曲線的形狀和性能。主要方法包括直接編碼和功能帶分析。</p><p><b>　　3.1直接編碼</b

74、></p><p>　　光譜曲線特征的重要思想是強(qiáng)調(diào)光譜曲線的形狀，所以直接編碼是一個(gè)非常方便的方法，更廣泛地使用二進(jìn)制編碼。其原理是在每一個(gè)像素帶比較值和屬性值分配代碼“0”或“1”，根據(jù)其價(jià)值。可以表示</p><p>　　在這里，是第i個(gè)波段的代碼，是這個(gè)公式的原始屬性值，T是比較值。一般來(lái)說，平均光譜向量，它也可以通過手動(dòng)的方法選擇，根據(jù)曲線的形狀，有時(shí)光譜向量的中位數(shù)可能使

75、用。</p><p>　　如果使用四元編碼，同一地區(qū)的比例是小于二進(jìn)制編碼，但在不同地區(qū)之間的比例大幅下降。所以四元的編碼是更有效地測(cè)量不同的像素之間的相似性。</p><p>　　因?yàn)橹苯泳幋a，將分散連續(xù)反照率成離散的代碼，編碼的結(jié)果是由閾值的影響明顯，將導(dǎo)致信息丟失。雖然它的操作非常簡(jiǎn)單，只用于一些應(yīng)用，要求精度低，閾值應(yīng)根據(jù)不同情況選擇。</p><p>　　

76、3.2光譜吸收或反射功能</p><p>　　所有頻段用于直接編碼不同，光譜吸收或反射的功能，只強(qiáng)調(diào)那些位于山谷或頂點(diǎn)的數(shù)值。這意味著在光譜曲線與當(dāng)?shù)刈畲蠡蜃钚〉臄?shù)值應(yīng)在第一，然后可以做進(jìn)一步的分析確定。在一般情況下，反照率是用來(lái)描述一個(gè)像素的屬性，因此與當(dāng)?shù)刈畲蟮膸Х瓷漤旤c(diǎn)，并與當(dāng)?shù)刈畹臀丈焦扔嘘P(guān)。</p><p>　　后來(lái)得到的位置及相關(guān)參數(shù)，可以做詳細(xì)的分析。一般使用兩種方法，一種

77、是直接編碼和分析功能帶，另一種是使用功能屬性和他們的參數(shù)計(jì)算一些量化指標(biāo)。</p><p>　　3.3光譜吸收或反射功能的編碼</p><p>　　功能組別的位置，直接用于光譜特征編碼。以下將采取吸收功能作為一個(gè)例子。如果一個(gè)功能帶是吸收谷的位置，它的代碼將是“1”，否則，其代碼為“0”。編碼完成后進(jìn)一步匹配，可以做比較。功能帶的位置，因?yàn)檫@些不確定性和高光譜成像過程中的錯(cuò)誤，也許在不久的

78、帶移動(dòng)，這將導(dǎo)致低匹配率。擴(kuò)展編碼方法，以減少帶位移的影響，提出和使用本文。其想法是，如果某樂隊(duì)的代碼是“1”，然后之前和它背后的樂隊(duì)將被分配了相同的代碼“1”，然后匹配和分析，將完成。</p><p>　　碼向量的相似性度量匹配位。配比，得到匹配帶總帶數(shù)的比例。在這項(xiàng)研究中，使用兩個(gè)比賽計(jì)劃。一個(gè)是匹配所有樂隊(duì)和另一只匹配那些功能組別的代碼。</p><p>　　基于上述分析，四項(xiàng)計(jì)劃的

79、使用和比較。它們是：（1）所有波段的直接編碼和匹配的所有波段；（2）直接編碼所有波段和配套功能帶；（3）所有波段擴(kuò)展編碼和匹配；（4）擴(kuò)展編碼和匹配功能帶。</p><p>　　從上面的分析和比較，光譜的吸收和反射特性，編碼和匹配，可以發(fā)現(xiàn)，雖然吸收和反射帶可以描述地物的光譜特性，應(yīng)使用有效的匹配操作，以克服噪聲的影響，功能帶位移和其他因素。在實(shí)際應(yīng)用中，吸收和反射，可用于提取專題信息，并有效地檢索某些類型的對(duì)象

80、。</p><p>　　光譜的吸收和反射特性的基礎(chǔ)上，吸收光譜指數(shù)（SA我）或光譜反射率（SR我）可以計(jì)算出波長(zhǎng)，功能波段反照率更等指標(biāo)，可以更精確地描述光譜特征在一些場(chǎng)合情況。</p><p>　　4光譜計(jì)算和轉(zhuǎn)換功能</p><p>　　兩者的相關(guān)性和相互補(bǔ)償中存在不同波段的高光譜遙感信息，所以許多新的功能可以得到一定的計(jì)算和組合功能隊(duì)并用于分類，信息提取和其他

81、任務(wù)。</p><p>　　4.1歸一化差異植被指數(shù)（NDVI）</p><p>　　在植被光譜應(yīng)用中起著非常重要的角色。它可以描述一些地物，如植被葉面積指數(shù)（，比植被和土壤，植被組件等信息。在某些分類（植被通常被用來(lái)作為一個(gè)獨(dú)立的分類功能。</p><p><b>　　4.2導(dǎo)數(shù)光譜</b></p><p>　　導(dǎo)數(shù)光

82、譜譜衍生技術(shù)也被稱為一階導(dǎo)線光譜。一個(gè)等級(jí)，并可以由公式計(jì)算兩個(gè)秩導(dǎo)數(shù)光譜。</p><p>　　使用類似上述的算法，可以計(jì)算每個(gè)階級(jí)衍生頻譜。衍生計(jì)算后結(jié)束時(shí)，我們可以發(fā)現(xiàn)，每種類型的地面物體可能有某些功能區(qū)別于其他實(shí)體在一定秩導(dǎo)數(shù)光譜，可用于識(shí)別信息。有時(shí)導(dǎo)數(shù)光譜圖像可以被用來(lái)作為直接輸入分類。雖然頻譜的衍生工具，可以提供新功能，除了原有的信息，將一些新的圖像后形成的衍生操作，將數(shù)據(jù)量急劇增加。形式秩導(dǎo)數(shù)光譜

83、，氮 - 2M帶將形成，所以如何處理的數(shù)據(jù)量和效率之間的關(guān)系成為一個(gè)新的問題。</p><p><b>　　5 結(jié)論和討論</b></p><p>　　在本文中，以光譜特征的高光譜遙感信息處理的需求為導(dǎo)向與光譜特征的框架，提出和一些主要的特征提取算法及其應(yīng)用進(jìn)行了討論，并完成了一些改進(jìn)，實(shí)驗(yàn)和分析。從本文的研究，可以得出以下結(jié)論：</p><p&g

84、t;　　1）在萃取的原理和算法的基礎(chǔ)上，高光譜遙感信息的光譜特征可以分為三個(gè)層次：光譜曲線特征，光譜變換和計(jì)算功能和光譜相似性度量功能。這個(gè)框架是用于進(jìn)一步的分析和應(yīng)用。</p><p>　　2）由于像素光譜向量的共同特性，某些功能可以提取和使用。二進(jìn)制編碼的算法和計(jì)算簡(jiǎn)單，容易，但它會(huì)導(dǎo)致一些細(xì)節(jié)信息的損失。第四紀(jì)編碼可以描述高光譜曲線特征用于匹配，檢索等方面的工作?；诠庾V曲線的反射??和吸收功能，有廣泛的應(yīng)

85、用，檢索，專題信息提取和其他任務(wù)，但必須采取有效匹配策略，以控制錯(cuò)誤。本文提出了兩個(gè)新的應(yīng)用程序，包括擴(kuò)展編碼和匹配，并結(jié)合反射和吸收特點(diǎn)匹配，得到詳細(xì)信息，比傳統(tǒng)方法在功能措施上的應(yīng)用更加簡(jiǎn)潔方便。</p><p>　　3）作為主要的計(jì)算和轉(zhuǎn)換功能，歸一化差異指數(shù)和導(dǎo)數(shù)光譜可以提供新的功能，參與分類，提取等加工和提取這些有用的模式和原始數(shù)據(jù)背后隱藏的信息，所以他們是非常方便的，在高光譜遙感信息處理。</p