多源遙感匯報(1)

1、1編輯ppt 2編輯ppt 1.研究區(qū)及數據 3編輯ppt 2.研究方法 面向對象的城市地物信息提 取方法主要包括個步驟：影 像分割、分類以及分類后處理。 進行多尺度分割，利用監(jiān)督 法分割精度評價方法選出各類 地物的最佳分割尺度，進行尺 度綜合獲得符合地物邊界的分 割結果，從而基于分割對象提 取航空影像上的光譜、紋理、 空間特征及LIDAR 數據中的高 程、強度等特征，結合 ReliefF 特征重要度進行特征選擇，最 后以多分類器組合方法進行分 類，并進行分類后處理。 4編輯ppt 2.1數據預處理 首先使用Terrascan將Lidar數據去除粗差，然后采用ENVI Lidar生成DTM，

2、波段計算得到nDSM(nDSMDSMDTM)，它記錄了所有高于地面物體(如建 筑物、植被等)的高程信息。然后將點云和影像配準，最后將正射影像及nDSM 進行波段合成。 2.2多尺度分割及尺度綜合 本文采用多尺度分水嶺影像分割算法，分割尺度設定為387（尺度間隔3）， 尺度為90以上基本合并為大區(qū)域，不予考慮，最終得到28個分割尺度。再利用 監(jiān)督法分割精度評價選擇各類地物的最佳分割尺度，并綜合最優(yōu)尺度，得到分 割結果。 準確度p：對象與分割對象的交集面 積與分割對象面積的比值 查全率r：則為交集面積與參考對象 面積的比值 m2：為交集面積與并集面積的比值 5編輯ppt 如圖5所示的 分割質量曲線

3、圖，其中建設用地包括建筑物、 道路、空地；林地包括樹木及草地。從圖中 可看出隨著分割尺度的增大，p值減少，r值 增加，過分割現象減少，欠分割現象增多， 而 m2為這兩種度量的加權和，不存在偏向 性。因此最佳分割尺度一般取m2最大值處， 若 m2近似則選取p值較大處，得出各類地物 類型的最佳分割尺度。樹木與草地為33，道 路與空地為45，陰影為18。 6編輯ppt 2.3對象分類 分類前，首先對每類地物選取其典型訓練樣本，統(tǒng)計對象光譜及指數特征、 紋理特征、形狀特征、對象高程信息、強度信息等40多種特征，如表1所示。 本文根據ReliefF指數進行特征重要性度量，順序選取前12項特征及面積、緊

4、湊度（用來區(qū)分道路與空地）進行影像分類。 7編輯ppt 特征選擇后采用兩步法提取城市地物，首先采用多分類器組合(K近鄰算法、 神經元網絡及SVM_RBF)對城市地物進行分類，得到如圖6所示的分類結果。由于 城市地區(qū)存在高層建筑及樹木，分類受陰影的影響較大，因此定本文在得出初步 分類結果后采用易康5.0將陰影區(qū)域進行分割，并基于規(guī)則將陰影區(qū)域依次分為： 陰影下植被、陰影下建筑物、陰影下道路、陰影下空地、純陰影5類。 圖7為A區(qū)最終分類結果，利用混淆矩陣法進行分類精度評價，總體精度為 93.1%，如表2所示 8編輯ppt 9編輯ppt 2.4 方法檢驗 為檢驗該方法的適用性，將該方法運用于B區(qū)，得

5、到B區(qū)分類結果，如圖8、 圖9所示。將其進行分類精度評價，如表3所示。實驗結果證明該方法能在保 證分類精度的前提下，推廣到類似研究區(qū)，有效提高分類效率，具有較高的 適用性，為大范圍城市地物信息快速提取提供了可能。 10編輯ppt 3 結束語 本文結合航拍影像與Lidar數據，利用監(jiān)督法分割精度評價方法從 多尺度分割結果中選擇城市典型地物最佳分割尺度，并進行尺度綜合， 能快速得到符合地物邊界的最優(yōu)分割結果。在提取高維影像及Lidar數 據的垂直結構特征基礎上，利用ReliefF特征重要度度量方法選擇最優(yōu) 特征組合，并結合多分類器組合方法進行城市地物信息提取，取得了 理想的分類效果。 基于監(jiān)督法分

6、割精度評價可以定量分析地物的最佳分割尺度，減 少了主觀判斷錯誤；基于ReliefF算法計算特征重要度則可以根據重要 度選擇分類特征，自動化程度提高；通過建立規(guī)則進行分類對象的后 處理將陰影區(qū)域進行細分，能有效降低陰影產生的分類誤差，該方法 在城市地物信息提取中具有較強的適用性，值得推廣。 不足：分割數據中nDSM高程誤差不可避免，由于光譜特征與高程特 征相同，對于植被而言，無法準確提取低矮樹木的輪廓，低矮數目與 草地無法分離，道路與空地分離效果也不是很好，需要進一步研究。 11編輯ppt 12編輯ppt Numerous coal fires burn underneath the Daton

7、g coalfield because of indiscriminate mining. Landsat TM/ETM,unmanned aerial vehicle (UAV), and infrared thermal imager were employed to monitor underground coal fires in the Majiliang mining area. The thermal field distributions of this area in 2000, 2002, 2006, 2007, and 2009 were obtained using L

8、andsat TM/ETM.The changes in the distribution were then analyzed to approximate the locations of the coal fires. Through UAV imagery employed at a very high resolution (0.2 m), the texture information, linear features, and brightness of the ground fissures in the coal fire area were determined. All

9、these data were combined to build a knowledge model of determining fissures and were used to support underground coal fire detection. An infrared thermal imager was used to map the thermal field distribution of areas where coal fire is serious. Results were analyzed to identify the hot spot trend an

10、d the depth of the burning point. 13編輯ppt The Datong coalfield is located in northern Shanxi Province, China approximately from 3952 to 4010north latitude and 1124932 to 113930 west longitude。 It has a complex terrain with an average altitude of more than 1200m. The studied area is located in the Ma

11、jiliang mining area Coal fires mainly occur in the No. 2 Jurassic coal seams in the mining area, 30m to 130m below the surface, with an average thickness of about 1m。 2 Studied area 14編輯ppt 3 Multi-source remote sensing monitoring the study was conducted on different scales： 1) Using the Landsat the

12、rmal infrared band, the temperature filed distribution of the entire study area was established and the whole area was classified according to its temperature. This procedure helped to determine roughly where the underground coal fires exist. 2) Using UAV technology, a high-resolution (0.2 m) image

13、of the studied area was obtained. Based on the characteristics of the ground fissures caused by underground coal fire, a new method was introduced to monitor underground coal fires. 3) For areas in which the underground coal fire was serious, a thermal infrared imager was employed to build a high-pr

14、ecision regional surface temperature distribution.The spatial variability of the temperature distribution was analyzed to determine the location under which points coal fires are burning. 15編輯ppt 1）At present, the main algorithms used in surface temperature retrieval include the radiative transfer e

15、quation, the mono-window algorithm, and the single-channel method。 2）Landsat ETM imageries in 2000 and 2002 and Landsat TM imageries in 2006, 2007, and 2009 were selected to obtain the surface temperature in each period (Fig. 2).Figure 2(a) shows the true color composite image of the studied area on

16、 September 25, 2002 at a resolution of 15m. The western part of the studied area is mainly composed of buildings, so that the corresponding areas in Figs. 2(b)(f) show red or yellow colors, which represent high surface temperature. The high temperature parts show a southwest to northeast trend, wher

17、e the temperature can be higher than the surrounding low temperature areas by around 10C to 30C. Such trend suggests that coal fires are burning underground。Landsat images help roughly determine target regions where underground coal fires exist. 16編輯ppt 17編輯ppt 18編輯ppt 19編輯ppt According to the textu

18、re information, linear feature,and brightness of the ground fissures, a knowledge model was established to facilitate the automatic extraction of ground fissures. The steps are as follows. 1)Occurrence-based variance, cooccurrence based variance, data range, and contrast filters are used for the UAV

19、 image to obtain the texture information in the fissured area by a moving window. 基于方差的發(fā)生與共生？ 2) Principle component analysis (PCA) and Fisher linear discrimination(線性 判別分析) analysis are then performed to extract the linear features of the area. 3) A gray level statistic is then made for the UAV ima

20、ge to obtain gray value samples of real fissures recorded by GPS. These characteristics are compared linear features of the area. 4) Based on the model established in Steps 1 to 3, ground fissures are automatically extracted using ERDAS software. Figure 5 shows a flowchart of the ground fissure extr

21、action using a UAV image and Fig. 6 shows the results. (b) Extracted fissures Fig. 6 20編輯ppt 3.3.1 Surface temperature field analysis of spontaneous combustion mountains In this study, a Th9100 Wri8.5 infrared thermal imager was used in implementing ground thermal infrared monitoring through a top-d

22、own parallel partition from the summit of the coal spontaneous combustion mountain and in collecting thermal infrared photos. The collected thermal infrared images were stitched in door according to their spatial coordinates. The results are shown in Fig7. As Fig8 shows, the surface temperature in t

23、he coal fire region of the Majiliang coal mining area exhibits a strong regularity in its spatial variation. The distribution of its high- temperature and low-temperature zones is relatively concentrated. The spatial distribution of the mountain surface temperature shows an obvious aggregation, whic

24、h indicates that deep and shallow layers of coal combustion zone exist underground. 21編輯ppt As the figures show, the temperature in the cracks is higher than that on the surrounding surface (Fig9). However, the temperature around the cracks is far lower than that around non-cracks, which shows that

25、the cracks indeed serve as channels of air exchange between the atmosphere and the coal combustion zone inside the mountain. The internal gases are discharged through the cracks, but they do not constitute a vertical form of the internal combustion point on the surface. 22編輯ppt As such, no one gener

26、al model can solve the problem, based on the study of the heat productionheat dissipation balance calculation equation of the spontaneous combustion depth of the inter gangue hill.This study modifies the equation to estimate depth of the ignition point of the underground coal fire.Considering that t

27、he studied area is mainly filled with coal combustion, the equation was modified by canceling the gangue carbon content correction value.The new coal combustion depth calculation equation is as follows: L=g(tfta)/(KKD ) According to the value of each parameter, the results are as follows. When the s

28、urface temperature is the highest (101.3 C), the burning depth of the underground coal fires is about 1.87 m. When the surface temperature is 0 C, the burning depth of the underground coal fires is about 7.33 m. When the surface temperature is equal to or lower than the air temperature,no spontaneou

29、s combustion phenomenon exists. 23編輯ppt On this basis, the logarithmic and power functions with a relatively high fitness are selected to fit the functions of the spontaneous combustion temperature and the depth of underground coal fire point, as shown in Fig10. The depth map of the internal spontan

30、eous combustion point is shown in Fig11. First, in the graph, the white region surrounded by the blue region is the region with the shallowest ignition depth. The white region surrounded by the red region is the region with the deepest ignition depth. Second, the calculation of underground coal combustion depth