論文(設(shè)計(jì))基于城市航空立體像對(duì)的全自動(dòng)三維建筑物建模07906
《論文(設(shè)計(jì))基于城市航空立體像對(duì)的全自動(dòng)三維建筑物建模07906》由會(huì)員分享,可在線閱讀,更多相關(guān)《論文(設(shè)計(jì))基于城市航空立體像對(duì)的全自動(dòng)三維建筑物建模07906(7頁(yè)珍藏版)》請(qǐng)?jiān)谘b配圖網(wǎng)上搜索。
1、專業(yè)好文檔 基于城市航空立體像對(duì)的全自動(dòng)三維建筑物建模收稿日期:2001-07-DD;截稿日期:2001-MM-DD 作者簡(jiǎn)介:陳愛(ài)軍(1972-),男(漢族),山西五寨人,清華大學(xué)計(jì)算機(jī)科學(xué)與技術(shù)系博士后,2000年7月畢業(yè)于北京大學(xué)遙感與GIS研究所,在國(guó)內(nèi)外會(huì)議刊物上已發(fā)表論文20多篇。主要研究方向?yàn)榱Ⅲw視覺(jué)建模、空間信息網(wǎng)絡(luò)共享與存取、數(shù)字城市、數(shù)字地球。 陳愛(ài)軍,徐光祐,史元春 (清華大學(xué)計(jì)算機(jī)科學(xué)與技術(shù)系 北京 100084) Automated 3D Building Modeling Based on Urban Aerial Stereo CHEN Ai-
2、jun, XU Guang-you, SHI Yuan-chun (Department of Computer Science and Technology, Tsinghua University, 100084) Abstract: Urban 3D building modeling is one of key technology of virtual city. In this paper, an approach to tackle the problem of 3D building modeling from urban high-resolution aerial
3、image is presented. Automatic 2D building detection technique and 3D height extraction technique has been applied to the test image (stereo) completely. Then combination between them is carried out, using the result of pyramidal stereo matching and the linear elements from 2D detection, the matching
4、 information from the matched point is assigned to the point of the linear element which corresponds to the matched point, so 3D modeling of building is achieved. In addition, more accurate 3D height information is obtained by applying new control strategies to prevent blunder propagation in pyramid
5、al matching which based on the modified ALSC algorithm. Key words: Virtual City, 2D Building Detection, Stereo Matching, Adaptive Least Square Correlation 摘要:城市建筑物三維建模是虛擬城市建設(shè)的關(guān)鍵技術(shù)之一。本文基于大比例尺航空立體像對(duì)提出三維建筑物建模方法。首先對(duì)二維邊緣檢測(cè)Canny算子進(jìn)行改進(jìn),以期從城市航空影像中檢測(cè)出較為精確的二維建筑物輪廓線。然后把改進(jìn)的ALSC算法運(yùn)用到金字塔匹配算法中,由于金字塔匹配算法中最為關(guān)鍵的
6、問(wèn)題是誤差傳遞的有效控制,而以往的研究中盡管采用了多種控制策略但效果不是很好,作者通過(guò)提出兩種新的誤差傳播控制策略得到了較高精度的匹配結(jié)果。利用匹配生成的高精度三維信息對(duì)檢測(cè)到的二維建筑物輪廓線進(jìn)行三維插值,獲得建筑物的三維信息,由此實(shí)現(xiàn)了建筑物三維建模。 關(guān)鍵詞:虛擬城市,二維建筑物檢測(cè),立體像對(duì)匹配,自適應(yīng)最小二乘相關(guān)(ALSC) 中圖分類號(hào):P23 0 引言 信息科技的發(fā)展證明:從計(jì)算機(jī)科學(xué)領(lǐng)域如計(jì)算機(jī)視覺(jué)和計(jì)算機(jī)圖形學(xué)到地球信息科學(xué)領(lǐng)域的攝影測(cè)量學(xué)科,一直到多學(xué)科領(lǐng)域交叉產(chǎn)生的WebGIS、數(shù)字城市和數(shù)字地球等,都對(duì)三維建筑物的建模與實(shí)現(xiàn)提出了迫切的需求。三維虛擬建筑物
7、的實(shí)現(xiàn)將對(duì)上述領(lǐng)域的前景產(chǎn)生前所未有的巨大影響,如三維虛擬城市、機(jī)器人智能導(dǎo)航、車輛輔助駕駛、建筑模擬展示、飛行模擬、交互式游戲、醫(yī)療模擬、虛擬購(gòu)物、虛擬博物館和虛擬藝術(shù)陳列館等。此外,還可應(yīng)用到計(jì)算機(jī)視覺(jué)領(lǐng)域視頻圖像的壓縮、瀏覽和檢索等以及數(shù)字?jǐn)z影測(cè)量領(lǐng)域的三維測(cè)繪等。所有這些應(yīng)用都將直接或間接地為數(shù)字地球建設(shè)提供解決現(xiàn)實(shí)世界三維數(shù)字化/虛擬化的有效途徑。 三維建筑物的國(guó)內(nèi)外研究可從地球信息科學(xué)的攝影測(cè)量學(xué)科與計(jì)算機(jī)科學(xué)的計(jì)算機(jī)視覺(jué)和計(jì)算機(jī)圖形學(xué)學(xué)科兩個(gè)方面進(jìn)行介紹。 數(shù)字?jǐn)z影測(cè)量學(xué)科中,主要研究基于地理空間矢量數(shù)據(jù)和城市大比例尺數(shù)字影像的三維城市建模與顯示。由于城市三維景物主要是人造建
8、筑物,所以三維建筑物信息的獲取與建模是城市三維建模的主要內(nèi)容。目前這方面的典型研究主要有:①?gòu)某鞘杏跋裰凶詣?dòng)提取建筑物,典型研究如檢測(cè)二維建筑物和DEM數(shù)據(jù)[1]、知覺(jué)組合[2]、線條分析[3]、使用陰影、透視幾何等輔助信息[4]、直接對(duì)建筑物或表面進(jìn)行建模[5]、基于知識(shí)的系統(tǒng)[6]以及通過(guò)影像測(cè)量并結(jié)合物體的幾何知識(shí)構(gòu)模出多面體對(duì)象模型的方法[7]等。②結(jié)合已有的二維地圖矢量數(shù)據(jù)利用航空激光掃描[8][9]或激光高度計(jì)數(shù)據(jù)[10][11]。③利用三維深度傳感器、多CCD相機(jī)和彩色高分辨率數(shù)字相機(jī)獲取的數(shù)據(jù)實(shí)現(xiàn)建筑物建模[12]。④利用虛擬現(xiàn)實(shí)(VR)技術(shù)實(shí)現(xiàn)三維GIS數(shù)據(jù)的可視化[13][
9、14]。其它方法如人機(jī)交互下的半自動(dòng)三維建筑物建模等[15][16]。 在計(jì)算機(jī)視覺(jué)和計(jì)算機(jī)圖形學(xué)中,主要研究既包括三維物體的建模和顯示,即從外向內(nèi)看的三維建模,又包括三維真實(shí)場(chǎng)景的建模和顯示,即從內(nèi)向外看的三維建模。目前方法分兩類:基于模型的繪制方法(MBR)和基于圖像的繪制方法(IBR)。①基于模型的繪制方法中,三維模型數(shù)據(jù)的獲取通常采用CAD的模型生成器或從實(shí)際環(huán)境中直接獲取。②基于圖片的繪制方法是通過(guò)一個(gè)來(lái)自多視點(diǎn)的原始的或合成的圖片庫(kù)來(lái)產(chǎn)生任意視點(diǎn)的新的虛擬圖片[17]。盡管對(duì)于復(fù)雜環(huán)境建模IBR技術(shù)優(yōu)于MBR技術(shù),但它能實(shí)際處理的三維對(duì)象范圍較小,對(duì)于大范圍城市建筑物建模目前還不
10、可行,而在采用MBR技術(shù)時(shí),城市建筑物三維數(shù)據(jù)的自動(dòng)獲取是研究重點(diǎn),尤其是城市密集區(qū)域三維數(shù)據(jù)的全自動(dòng)獲取。本文是結(jié)合二維圖像建筑物的自動(dòng)提取和三維高度信息的自動(dòng)產(chǎn)生來(lái)解決該問(wèn)題的。 1 改進(jìn)的ALSC算法 自適應(yīng)最小二乘相關(guān)(Adaptive Least Square Correlation, ALSC)算法[17]是瑞士聯(lián)邦理工大學(xué)的Gruen教授提出。假定立體像對(duì)的左、右影像分別有灰度級(jí)函數(shù)f (x, y)和g (x, y),如果左右影像中的一對(duì)共軛點(diǎn)(x0, y0)和(x0’, y0’)理想相關(guān),則有: f (x0, y0) = g (x0’, y0’) (1) 但由于左
11、右影像中隨機(jī)噪聲的影響(或者假定只有匹配圖片有噪聲,而模板沒(méi)有噪聲)(我們把左影像稱為匹配模板,右影像稱為匹配圖片),所以等式(1)通常不成立。給匹配圖片增加一個(gè)噪聲矢量(含量測(cè)噪聲和模型誤差)并對(duì)左右影像上所有點(diǎn)一般化,則有: f (x, y) – e (x, y) = g (x, y) (2) 根據(jù)最小二乘理論,等式(2)是非線性觀察等式,它用函數(shù)g (x, y)表示觀察矢量f (x, y),而函數(shù)g (x, y)在右影像中的位置需要估計(jì),此位置可通過(guò)相對(duì)于函數(shù)g (x, y)的初始位置,即共軛圖片區(qū)域的近似值g0 (x, y)的偏移參數(shù)△x、△y來(lái)表示。 為獲得較好的匹配結(jié)果,
12、考慮到影像本身的各種畸變,除采用偏移參數(shù)外,還引入影像變形參數(shù)和輻射校正。 假定與f (x, y)匹配的共軛點(diǎn)是g0 (x, y),那么發(fā)生幾何變形的匹配圖片上真正和f (x, y)匹配的共軛點(diǎn)可通過(guò)對(duì)g0 (x, y)的二變量高階多項(xiàng)式的變換得到,即: x = tyTAtx (3a) y = tyTBtx (3b) 其中txT = {1,x0, x02, ···x0m-1}; tyT = {1,y0, y02, ···y0m-1} 參數(shù)矩陣A,B: b11 b12 ··· b1m bm1 bm2 ··· bmm B = a11 a12
13、··· a1m am1 am2 ··· amm A = x0,y0是匹配圖片g0 (x, y)數(shù)據(jù)點(diǎn)的網(wǎng)格位置。 變換參數(shù)a11 ··· amm,b11··· bmm需從(2)中估計(jì)得出。 g0 (x, y) x 乙乙乙乙乙乙丶 · 乙丶x g0 (x, y) 為使用常規(guī)的最小二乘方法處理等式(2),用泰勒公式對(duì)等式(2)的右邊進(jìn)行線性變換: dy 乙乙乙乙乙乙丶 · 乙丶x f (x, y) – e (x, y) = g0 (x, y) + + dx+
14、 乙乙乙乙乙乙丶 · 乙丶x (4) 乙乙乙乙乙乙丶 · 乙丶x x 乙乙乙乙乙乙丶 · 乙丶x dy = 乙乙乙乙乙乙丶 · 乙丶x y pi 乙乙乙乙乙乙丶 · 乙丶x dpi 乙乙乙乙乙乙丶 · 乙丶x dx = 乙乙乙乙乙乙丶 · 乙丶x x pi 乙乙乙乙乙乙丶 ·
15、 乙丶x dpi 乙乙乙乙乙乙丶 · 乙丶x 其中 pi為(3)中第i個(gè)變換參數(shù) 對(duì)參數(shù)矩陣A、B,我們僅取前兩行兩列,并令二次項(xiàng)的系數(shù)為零,即: a11 a12 a21 0 A = b11 b12 b21 0 B = (5) 乙乙乙乙乙乙丶 · 乙丶x x = a11 + a12x0 + a21y0 y = b11 + b12x0 + b21y0 乙乙乙乙乙乙丶 · 乙丶x (6) 乙乙乙乙乙乙
16、丶 · 乙丶x 這樣由(3)可得變換: (6)考慮了匹配圖片完整的仿射影像變形,同時(shí)也包含了偏移參數(shù)△x、△y,此處用a11、b11表示。對(duì)(6)微分得: dx = da11 + x0da12 + y0da21 dy = db11 + x0db12 + y0db21 乙乙乙乙乙乙丶 · 乙丶x (7) 乙乙乙乙乙乙丶 · 乙丶x g0(x, y) y 乙乙乙乙乙乙丶 · 乙丶x gy = 乙乙乙乙乙乙丶 ·
17、 乙丶x 采用簡(jiǎn)化形式: g0(x, y) x 乙乙乙乙乙乙丶 · 乙丶x gx = 乙乙乙乙乙乙丶 · 乙丶x (8) 乙乙乙乙乙乙丶 · 乙丶x 將(7)(8)代入(4),并增加輻射偏移參數(shù)rs得: f (x, y) – e (x, y) = g0 (x, y) + gxda11 + gxx0da12 + gxy0da21 + gydb11 + gyx0db12 + gyy0db21 + rs XT = {da11 da12 da2
18、1 db11 db12 db21 rs}為參數(shù)矢量 A = {gx gxx0 gxy0 gy gyx0 gyy0 1}為設(shè)計(jì)矩陣 b = f (x, y) - g0 (x, y)為模板矢量和圖片矢量的差 ε= e (x, y)為噪聲誤差 (9) 乙乙乙乙乙乙丶 · 乙丶x 令: 則(9)變?yōu)椋? AX = b –ε (10) 假定噪聲ε(i)是同分布的,獨(dú)立的且具有零均值及方差為δ2(ε具有白噪聲性質(zhì),是一個(gè)具有零均值的平穩(wěn)隨機(jī)矢量)即: E(ε) = 0, var(ε) = E(εεT) = δ2I (
19、11) (10)(11)共同組成Gauss-Markov估計(jì)模型,該模型指出當(dāng)噪聲是白色時(shí),它的最小二乘估計(jì)是無(wú)偏的、有效的和一致的。由該模型可得參數(shù)矢量的最小二乘估計(jì)值為: X’ = (ATWA)-1ATWb (12) 其中W為由加到每個(gè)誤差項(xiàng)εi的不同的權(quán)組成的權(quán)重矩陣,所以(12)是加權(quán)最小二乘估計(jì)。當(dāng)對(duì)每個(gè)誤差項(xiàng)εi加相同的權(quán)時(shí),W = I,即權(quán)重矩陣變?yōu)閱挝痪仃?。這樣(12)簡(jiǎn)化為: X’ = (ATA)-1ATb (13) 上式稱為普通最小二乘估計(jì)。由于實(shí)際問(wèn)題(2)的非線性,最終解可通過(guò)(12)或(13)迭代求出。 設(shè)初始迭代參數(shù)近
20、似為: a011 = b011 = a021 = b012 = 0, a012 = b021 = 1 則初始估計(jì)值坐標(biāo)集合為: xi = x0i , yi = y0i , i = 1, 2, ….n 其中n為匹配模板或匹配圖片中的格網(wǎng)點(diǎn)個(gè)數(shù)。 由(13)求得參數(shù)矢量后,應(yīng)用(6)進(jìn)行變換,g0(x, y)可在變換后得到的新坐標(biāo)上重新計(jì)算,得到新的系數(shù)矩陣A。再利用(13)求新的參數(shù)矢量,如此迭代。直到參數(shù)矢量中每個(gè)參數(shù)的變化都小于某一特定值,迭代停止,求得最終的參數(shù)矢量的解。用此解代入(6)可在匹配圖片中求得與匹配模板中匹配點(diǎn)精確匹配的共軛點(diǎn)。 ALSC算法只能匹配左右影
21、像中的一對(duì)共軛點(diǎn),不能匹配所有點(diǎn)。文獻(xiàn)[18]通過(guò)用于影像分割的區(qū)域增長(zhǎng)算法擴(kuò)展了ALSC算法,使得ALSC算法可用于匹配左右影像的全部點(diǎn)。算法如下: INPUTS: two image; 1 or more approximate matches between the images Set list_matched_point to empty For each approximate match Run ALSC’ algorithm If it converges Store result to list_matched_point While l
22、ist_matched_point is not empty Pick an item from the list (and remove it) For each ‘neighbour’ of the selected match If ‘neighbour’ not already match Use selected item to predict match Run ALSC’s algorithm using prediction If it converges (and satisfies any constraints we might i
23、mpose) Store result in list_matched_point 輸出結(jié)果可以是計(jì)算的直接結(jié)果,也可以是其它使用方便的定義形式。匹配點(diǎn)鄰域選取影像網(wǎng)格化后的格網(wǎng)中與該匹配點(diǎn)(格網(wǎng)點(diǎn))最近的四個(gè)點(diǎn)。 2 建筑物三維信息自動(dòng)獲取 立體像對(duì)相關(guān)匹配是獲取三維數(shù)據(jù)的主要方法之一。常用的立體匹配方法主要有基于區(qū)域的、基于特征的和基于相位的匹配,本文采用基于區(qū)域的結(jié)合改進(jìn)型ALSC算法的金字塔匹配方法。 金字塔匹配方法既有優(yōu)點(diǎn)又有缺點(diǎn),優(yōu)點(diǎn)在于可加快匹配速度,并可極小化阻礙自動(dòng)立體匹配的某些特征,如SAR影像中的裂紋和城市航片中的不連續(xù)。缺點(diǎn)是容易造成誤差傳播,低層誤差
24、會(huì)傳播到高層,并一層層放大,最終可能導(dǎo)致匹配結(jié)果不可用。圖2是結(jié)合改進(jìn)的ALSC算法的金字塔匹配算法流程圖。 在金字塔匹配的每一層都需要種子點(diǎn),最低層(即分辨率最低的層)的初始種子點(diǎn)需要明確給出,較高層的匹配需要選擇性地使用低層的匹配結(jié)果。為有效控制誤差的傳播,用于高層匹配的來(lái)自于低層匹配結(jié)果的種子點(diǎn)選取策略對(duì)金字塔匹配算法的成功與否是至關(guān)重要的。為提高立體匹配性能,不好的匹配點(diǎn)應(yīng)盡量排除,不作為高層匹配的種子點(diǎn),但種子點(diǎn)又必須盡量分布在整幅影像中。因此,種子點(diǎn)的選取策略將直接影響到金字塔匹配算法的有效性和實(shí)用性。本文中作者提出如下兩種選取策略。 l 基于精度的策略:用最小二乘估計(jì)模型的估
25、計(jì)誤差的協(xié)方差陣的最大特征值作為匹配精度的度量值。估計(jì)誤差的協(xié)方差矩陣為: Cov(X’ - X) = δ2(ATA)-1 其中δ2 可根據(jù) var(ε) = E(εεT) = δ2I求得。 預(yù)定義一個(gè)閾值(50),匹配精度小于閾值的低層匹配點(diǎn)將作為高層匹配的種子點(diǎn)。 l 基于瓦片的策略:把低層影像分成許多小瓦片,僅選擇每個(gè)瓦片中精度最好的匹配點(diǎn)作為高層匹配種子點(diǎn)。瓦片大小的選擇依賴于匹配影像中最小建筑物的像素大小,我們選取的5層金字塔上每一層的瓦片大小由低層到高層分別為3×3、5×5、9×9、18×18和35×35像素。 在實(shí)驗(yàn)過(guò)程中,基于精度的選取策略存在的問(wèn)題是選取
26、的種子點(diǎn)不能均勻分布在整幅影像上,而在基于瓦片的選取中,由于在整幅影像上均勻分布地定義瓦片,使得種子點(diǎn)遍布整幅影像。二者的結(jié)合獲得了精度較高的種子點(diǎn)。 在文獻(xiàn)[1]的闡述中僅簡(jiǎn)單地提到了采用基于瓦片的過(guò)濾技術(shù),具體策略及結(jié)果如何并沒(méi)有詳細(xì)論述。本文利用文獻(xiàn)[17]的數(shù)學(xué)推理,在文獻(xiàn)[1]的基礎(chǔ)上提出新的誤差傳播控制策略,得到了較好的匹配精度。 我們選取金字塔層數(shù)為5層,最低層種子點(diǎn)選取通過(guò)假定一個(gè)視差常數(shù)并使用預(yù)匹配來(lái)非隨機(jī)選取。假定影像的視差范圍為30-200像素,那么在最低層的視差范圍是1-6像素。同時(shí)假定最低層視差常數(shù)為零,由此在左右影像中按照格網(wǎng)點(diǎn)產(chǎn)生視差為零的種子點(diǎn),那么在右影像
27、中的實(shí)際共軛點(diǎn)將落在以初始點(diǎn)為中心的6像素視差范圍內(nèi)。高層種子點(diǎn)從低層獲取,這樣可實(shí)現(xiàn)全自動(dòng)立體像對(duì)匹配。對(duì)匹配結(jié)果應(yīng)用攝像機(jī)模型轉(zhuǎn)換為地面坐標(biāo),得到建筑物三維信息,建筑物高度信息可保存為規(guī)則格網(wǎng)的數(shù)字高程模型(DEM)。 3 二維建筑物檢測(cè) 對(duì)大比例尺航空相片進(jìn)行二維建筑物檢測(cè)需要用到計(jì)算機(jī)視覺(jué)的低層技術(shù)。本文提出基于關(guān)系圖的檢測(cè)方法的流程結(jié)構(gòu)圖如圖2。 d = fb α+ β 原始影像 邊緣檢測(cè) 邊緣長(zhǎng)度和方向 線性擬合 線 線關(guān)系搜索 線關(guān)系圖 生成候選建筑物 驗(yàn)證候選建筑物 二維建筑物輪廓 圖2、二維建筑物檢測(cè)流程圖 Fig.2 A sche
28、matic flow-chart for 2D building detection 1層 0層 ······ 左影像 航片立體匹配 左影像 初始種子點(diǎn)產(chǎn)生 第0層匹配結(jié)果 左影像 航片立體匹配 左影像 第1層匹配結(jié)果 左影像 航片立體匹配 左影像 最后匹配點(diǎn)輸出 種子點(diǎn)選取策略 種子點(diǎn) 種子點(diǎn) 種子點(diǎn)選取策略 攝像機(jī)模型使用 DEM生成 圖1 金字塔匹配流程圖 Fig.1 Schematic flow-chart of pyramidal matching n層 選取立體像對(duì)中的左影像,用5×5像素的CPK[19][20]過(guò)濾器進(jìn)行邊
29、緣檢測(cè),得到邊緣強(qiáng)度和邊緣方向。用Hough變換和連結(jié)邊緣標(biāo)注方法從邊緣中檢測(cè)出線元素,并用端點(diǎn)模板搜索線元素的兩個(gè)端點(diǎn)。線元素用兩個(gè)端點(diǎn)和最小線長(zhǎng)(根據(jù)特定場(chǎng)景預(yù)定義的閾值)來(lái)定義,并合并與較長(zhǎng)的線斷開(kāi)的短線段和位置緊鄰的平行線。 這兩種方法都較好地?cái)M合出了建筑物邊線,將兩種方法各自產(chǎn)生的結(jié)果取交集,可得到更加可靠的建筑物邊線。 對(duì)所得建筑物邊線進(jìn)行線關(guān)系檢索,并存儲(chǔ)為線關(guān)系圖。為減小關(guān)系圖,長(zhǎng)度小于預(yù)定義閾值的線的關(guān)系和與其它線的夾角不近似于直角的線的關(guān)系將被刪除。用深度優(yōu)先算法遍歷關(guān)系圖搜索線關(guān)系圖中的閉合環(huán)生成候選建筑物。在候選建筑物驗(yàn)證中合并相似候選建筑物,并根據(jù)預(yù)定義閾值刪除平
30、均高度小于閾值的由建筑物地面基線構(gòu)成的候選建筑物(也可用陰影分析和透視幾何等輔助方法或考慮在此時(shí)引入DEM數(shù)據(jù))。最終得到建筑物二維輪廓線。 4 三維建筑物自動(dòng)生成 我們選取清華大學(xué)1987年1:8000校園航空立體像對(duì)的一部分(如圖3)用于二維建筑物檢測(cè)和立體匹配實(shí)驗(yàn)。 圖3、 實(shí)驗(yàn)用清華大學(xué)航空立體像對(duì) Fig.3 Aerial stereo image of Tsinghua Univ. for experiment 攝像機(jī)模型采用理想模型: 其中d為所求建筑物高度信息,α+ β為立體匹配產(chǎn)生的視差信息,f為攝像機(jī)透鏡的焦距,b為拍攝兩幅航片時(shí)攝像機(jī)透鏡的焦心移動(dòng)的水平距離。
31、采用在整個(gè)校園航片上大致均勻分布的多個(gè)地面控制點(diǎn)可確定建筑物高度和視差之間的比例關(guān)系,由此根據(jù)匹配結(jié)果(即視差)求出所有匹配對(duì)應(yīng)點(diǎn)的高度信息,進(jìn)而確定匹配對(duì)應(yīng)點(diǎn)的三維地面坐標(biāo),生成DEM數(shù)據(jù)。目前獲得的DEM數(shù)據(jù)精度可達(dá)到3米左右,對(duì)于更大比例尺的航片可望得到更高的精度,所以立體匹配中關(guān)鍵是匹配的像素精度能達(dá)到多少,目前可達(dá)到亞像素(例如假定1像素對(duì)應(yīng)地面10米,那么達(dá)到亞像素級(jí)后誤差還將有5米左右)。此外,建筑物輪廓線的提取精度達(dá)到了80%以上。 用立體匹配生成的DEM數(shù)據(jù)對(duì)二維建筑物檢測(cè)出的建筑物輪廓范圍進(jìn)行插值,如果某匹配點(diǎn)對(duì)應(yīng)于候選建筑物輪廓內(nèi)的某點(diǎn),則把匹配點(diǎn)DEM信息賦值給輪廓內(nèi)
32、的點(diǎn)。通過(guò)這種簡(jiǎn)單的二維建筑物輪廓信息和三維高度信息的融合可得到城市航空相片中建筑物的三維信息。對(duì)由于遮擋而不能獲得的建筑物信息通過(guò)建筑物的對(duì)稱性和邊緣的直線性可計(jì)算求得。建筑物形狀用簡(jiǎn)單的結(jié)構(gòu)如三角形、長(zhǎng)方形、棱形、錐形和園屋頂組成。這樣,可實(shí)現(xiàn)城市建筑物的全自動(dòng)三維建模。 5 結(jié) 論 實(shí)驗(yàn)證明對(duì)于表面近似光滑和視差近似連續(xù)的城市航空相片,把改進(jìn)的ALSC算法運(yùn)用到金字塔匹配算法中,采用初始種子點(diǎn)控制策略和誤差傳播控制策略,不僅可實(shí)現(xiàn)從建筑物二維信息檢測(cè)和立體匹配的全自動(dòng)三維信息獲取與建模,而且可得到亞像素級(jí)的匹配精度,同時(shí)立體匹配速度大大加快。只要有效控制金字塔匹配中低層到高層的誤差傳
33、播,就可大大提高匹配精度。因此,高精度快速的城市航空影像中建筑物三維信息的全自動(dòng)提取和建模得到了較好的解決。 進(jìn)一步需要研究的問(wèn)題在于盡管取得了亞像素級(jí)匹配精度,但精度還是遠(yuǎn)遠(yuǎn)不夠。考慮用最小二乘理論的擴(kuò)展理論進(jìn)一步提高估計(jì)參數(shù)精度,期望像素匹配精度達(dá)到0.2像素。 參考文獻(xiàn) 1 Kim T, Muller J. A Technique for 3D Building Reconstruc-tion. Photogrammetric Engineering & Remote Sensing, 1998, 64(9): 923-930. 2 Mohan R, Nevatia R. Usi
34、ng Perceptual Organization to Extract 3D Structures. IEEE Trans On Pattern Recognition and Machine Intelligence, 1989, 11(11): 1121-1139. 3 Shuflet , Mckeown D M. Fusion of Monocular Cues to Detect Man-Made Structure in Aerial Imagery. CVG&IP: Image Understanding, 1993, 57(3): 307-330. 4 Huertas A
35、, Nevatia R. Detecting Building in Aerial Images. CVG&IP, 1988, 41:131-152. 5 Huertas A, Bejanin M, Nevatia R. Model Registration and Validation. Automated Extraction of Man-Made Objects from Aerial and Space Images, Birkhauser, 1996: 33-42. 6 Nicolin B, Gabler R. A Knowledge-Based System for the
36、Analysis and Aerial Images. IEEE Trans On Geoscience and Remote Sensing, 1987, GE-25(3): 317-329. 7 Heuvel F A. 3D reconstruction from a single image using geometric constraints. ISPRS J. of Photogram. & Remote Sensing 1998, 53: 354-368. 8 Wehr A, Lohr U. Airbone laser scanning – an introduction a
37、nd overview. ISPRS J. of Photogram. & Remote Sensing 1999, 54: 68-82. 9 Ackermann F. Airbone laser scanning – present status and future expectations. ISPRS J. of Photogram. & Remote Sensing 1999, 54: 64-67. 10 Haala N, Brenner C. Virtual City Models from Laser Altimeter and 2D Map Data. Photogramm
38、etric Engineering & Remote Sensing, 1999, 65(79): 787-795. 11 Maas H G, Vosselman G. Two algorithms for extracting building models from raw laser altimetry data. ISPRS J. of Photogram. & Remote Sensing 1999, 54 (2-3): 153-163. 12 El-Hakim S F, Brenner C, G Roth. A multi-sensor approach to creatin
39、g accurate virtual environments. ISPRS J. of Photogram. & Remote Sensing 1998, 53: 379-391. 13 Germs R, Maren G V, Verbree E, F W Jansen. Virtual Reality & 3D GIS: A mutli-view VR interface for 3D GIS. Computers & Graphics, 1999, 23: 497-506. 14 Huang B, Lin H. GeoVR: a web-based tool for virtual
40、reality presentation from 2D GIS data. Computer & Geosciences, 1999, 25: 1167-1175. 15 Dai C G, Gong D C. Aerial Photograph Based Technology for Generation of Urban 3D Landscape. Bulletin of surveying and mapping, 2000, (10): 28-31. (in Chinese). 16 Zhang J Q, Zhang Z X, Xu F. Reconstruction of 3D
41、 Landscape from Images with Large Scale for Urban Region. J. of Wuhan Technogical Univ. of Surveying and Mapping, 1998, (12): 355-358. (in Chinese). 17 Greun A W. Adaptive Least Square Correlation: A Powerful Image Matching Technique. S Afr J of Photogram., Remote Sensing and Cartography, 1985, 14(
42、3): 175-187. 18 Otto G P, Chau T K. A Region-Growing Algorithm for Matching of Terrain Images. Image and Vision Computing, 1989, 7(2): 83-94. 19 Canny J. A Computational Approach to Edge Detection. IEEE Trans on Pattern Analysis and Machine Intelligence, PAMI1986, 8(6): 679-697. 20 Petrou M, Kitt
43、ler J. Optimal Edge Detection for Ramp Edges. IEEE Trans on Pattern Analysis and Machine Intelligence, 1991, 13(5): 483-491. Editor's note: Judson Jones is a meteorologist, journalist and photographer. He has freelanced with CNN for four years, covering severe weather from tornadoes to typhoons. Fo
44、llow him on Twitter: @jnjonesjr (CNN) -- I will always wonder what it was like to huddle around a shortwave radio and through the crackling static from space hear the faint beeps of the world's first satellite -- Sputnik. I also missed watching Neil Armstrong step foot on the moon and the first sp
45、ace shuttle take off for the stars. Those events were way before my time. As a kid, I was fascinated with what goes on in the sky, and when NASA pulled the plug on the shuttle program I was heartbroken. Yet the privatized space race has renewed my childhood dreams to reach for the stars. As a mete
46、orologist, I've still seen many important weather and space events, but right now, if you were sitting next to me, you'd hear my foot tapping rapidly under my desk. I'm anxious for the next one: a space capsule hanging from a crane in the New Mexico desert. It's like the set for a George Lucas movi
47、e floating to the edge of space. You and I will have the chance to watch a man take a leap into an unimaginable free fall from the edge of space -- live. The (lack of) air up there Watch man jump from 96,000 feet Tuesday, I sat at work glued to the live stream of the Red Bull Strat
48、os Mission. I watched the balloons positioned at different altitudes in the sky to test the winds, knowing that if they would just line up in a vertical straight line "we" would be go for launch. I feel this mission was created for me because I am also a journalist and a photographer, but above all
49、 I live for taking a leap of faith -- the feeling of pushing the envelope into uncharted territory. The guy who is going to do this, Felix Baumgartner, must have that same feeling, at a level I will never reach. However, it did not stop me from feeling his pain when a gust of swirling wind kicked u
50、p and twisted the partially filled balloon that would take him to the upper end of our atmosphere. As soon as the 40-acre balloon, with skin no thicker than a dry cleaning bag, scraped the ground I knew it was over. How claustrophobia almost grounded supersonic skydiver With each twist, you could
51、see the wrinkles of disappointment on the face of the current record holder and "capcom" (capsule communications), Col. Joe Kittinger. He hung his head low in mission control as he told Baumgartner the disappointing news: Mission aborted. The supersonic descent could happen as early as Sunday. The
52、 weather plays an important role in this mission. Starting at the ground, conditions have to be very calm -- winds less than 2 mph, with no precipitation or humidity and limited cloud cover. The balloon, with capsule attached, will move through the lower level of the atmosphere (the troposphere) whe
53、re our day-to-day weather lives. It will climb higher than the tip of Mount Everest (5.5 miles/8.85 kilometers), drifting even higher than the cruising altitude of commercial airliners (5.6 miles/9.17 kilometers) and into the stratosphere. As he crosses the boundary layer (called the tropopause), he
54、 can expect a lot of turbulence. The balloon will slowly drift to the edge of space at 120,000 feet (22.7 miles/36.53 kilometers). Here, "Fearless Felix" will unclip. He will roll back the door. Then, I would assume, he will slowly step out onto something resembling an Olympic diving platform. Be
55、low, the Earth becomes the concrete bottom of a swimming pool that he wants to land on, but not too hard. Still, he'll be traveling fast, so despite the distance, it will not be like diving into the deep end of a pool. It will be like he is diving into the shallow end. Skydiver preps for the big ju
56、mp When he jumps, he is expected to reach the speed of sound -- 690 mph (1,110 kph) -- in less than 40 seconds. Like hitting the top of the water, he will begin to slow as he approaches the more dense air closer to Earth. But this will not be enough to stop him completely. If he goes too fast or
57、spins out of control, he has a stabilization parachute that can be deployed to slow him down. His team hopes it's not needed. Instead, he plans to deploy his 270-square-foot (25-square-meter) main chute at an altitude of around 5,000 feet (1,524 meters). In order to deploy this chute successfully,
58、he will have to slow to 172 mph (277 kph). He will have a reserve parachute that will open automatically if he loses consciousness at mach speeds. Even if everything goes as planned, it won't. Baumgartner still will free fall at a speed that would cause you and me to pass out, and no parachute is g
59、uaranteed to work higher than 25,000 feet (7,620 meters). It might not be the moon, but Kittinger free fell from 102,800 feet in 1960 -- at the dawn of an infamous space race that captured the hearts of many. Baumgartner will attempt to break that record, a feat that boggles the mind. This is one of those monumental moments I will always remember, because there is no way I'd miss this.
- 溫馨提示:
1: 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
2: 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
3.本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒(méi)有圖紙預(yù)覽就沒(méi)有圖紙。
4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
5. 裝配圖網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 6.煤礦安全生產(chǎn)科普知識(shí)競(jìng)賽題含答案
- 2.煤礦爆破工技能鑒定試題含答案
- 3.爆破工培訓(xùn)考試試題含答案
- 2.煤礦安全監(jiān)察人員模擬考試題庫(kù)試卷含答案
- 3.金屬非金屬礦山安全管理人員(地下礦山)安全生產(chǎn)模擬考試題庫(kù)試卷含答案
- 4.煤礦特種作業(yè)人員井下電鉗工模擬考試題庫(kù)試卷含答案
- 1 煤礦安全生產(chǎn)及管理知識(shí)測(cè)試題庫(kù)及答案
- 2 各種煤礦安全考試試題含答案
- 1 煤礦安全檢查考試題
- 1 井下放炮員練習(xí)題含答案
- 2煤礦安全監(jiān)測(cè)工種技術(shù)比武題庫(kù)含解析
- 1 礦山應(yīng)急救援安全知識(shí)競(jìng)賽試題
- 1 礦井泵工考試練習(xí)題含答案
- 2煤礦爆破工考試復(fù)習(xí)題含答案
- 1 各種煤礦安全考試試題含答案