唐 杰,張 巖,范聰慧,程曉鑫,鄧家勇
?
使用高分遙感立體影像提取黃土丘陵區(qū)切溝參數(shù)的精度分析
唐 杰,張 巖※,范聰慧,程曉鑫,鄧家勇
(北京林業(yè)大學(xué)水土保持學(xué)院,水土保持國(guó)家林業(yè)局重點(diǎn)實(shí)驗(yàn)室,北京 100083)
為了研究高分立體像對(duì)測(cè)量黃土丘陵區(qū)切溝參數(shù)的適用性,選取陜北黃土區(qū)合溝小流域,以三維激光掃描全站儀獲取的數(shù)據(jù)為參照值,分析使用GeoEye-1高分遙感立體像對(duì)測(cè)量切溝參數(shù)的精度,得到如下研究結(jié)果。1)切溝面積、周長(zhǎng)、溝長(zhǎng)和溝寬等線狀和面狀參數(shù)平均測(cè)量誤差分別為3.58 m2,0.55 m,0.13 m和-0.10 m,其中面積、周長(zhǎng)和溝長(zhǎng)的百分誤差主要集中在5%以內(nèi),溝寬百分誤差主要分布在10%以內(nèi)。2)切溝三維參數(shù)溝底寬、最大溝深、平均溝深的平均測(cè)量誤差分別為-0.67、0.14和-0.46 m。截面積和體積的平均誤差分別為-6.30 m2和-54.01 m3。最大溝深的百分誤差主要集中在30%以內(nèi),溝底寬、平均溝深、截面積和體積的百分誤差則主要分布在50%以內(nèi);相較于三維激光掃描的切溝,立體像對(duì)提取的切溝溝底形態(tài)誤差較大,主要是溝底寬和平均溝深偏小。3)切溝規(guī)模越大,切溝體積、截面積和溝底寬的測(cè)量值偏小的幅度越大。但是,切溝體積測(cè)量誤差與切溝體積之間可以建立較好的線性回歸模型,在缺少其他測(cè)量手段時(shí),可以使用該模型對(duì)測(cè)量誤差進(jìn)行校正??傮w上看,高分立體遙感為切溝線狀和面狀參數(shù)測(cè)量以及切溝體積測(cè)量提供了新的方法,為黃土丘陵區(qū)溝蝕監(jiān)測(cè)提供了便捷、且相對(duì)可靠的數(shù)據(jù)源。
遙感;測(cè)量;提??;3D激光掃描;切溝形態(tài)參數(shù);測(cè)量誤差;黃土丘陵區(qū)
黃土高原因其獨(dú)特的地貌和脆弱的生態(tài)環(huán)境成為世界上水土流失最為嚴(yán)重的地區(qū)之一[1-3]。溝蝕是黃土高原水土流失的主要形式之一,切溝侵蝕產(chǎn)沙量占流域產(chǎn)沙量的50%以上[4],由于切溝發(fā)育的階段性和切溝形態(tài)的復(fù)雜性以及監(jiān)測(cè)手段不足,使得較大時(shí)空尺度的切溝參數(shù)測(cè)量極其困難,從而嚴(yán)重制約了切溝侵蝕機(jī)理和切溝侵蝕預(yù)報(bào)研究[5-6]。因此,采用不同手段對(duì)黃土高原切溝進(jìn)行持續(xù)監(jiān)測(cè)并探索其變化規(guī)律與影響因素十分必要[7-8]。
近年來(lái),多種遙感手段在切溝侵蝕監(jiān)測(cè)研究中取得了較大進(jìn)展。數(shù)字高程模型(DEM)廣泛應(yīng)用于土壤侵蝕、地貌變化等地形分析[9-10],胡剛等[11]利用GPS采集的點(diǎn)數(shù)據(jù)生成DEM提取地貌數(shù)據(jù),探索GPS應(yīng)用于溝蝕研究的可行性;Wu等[12]根據(jù)GPS采集數(shù)據(jù)計(jì)算綏德小流域切溝形態(tài)參數(shù)并監(jiān)測(cè)其侵蝕速率。三維激光掃描技術(shù)與激光雷達(dá)技術(shù)可以快速獲取所測(cè)對(duì)象的三維地形數(shù)據(jù),并最終通過(guò)獲取該區(qū)域的高分辨率DEM對(duì)象進(jìn)行動(dòng)態(tài)監(jiān)測(cè)[13-15]。雖然三維激光掃描儀具有接觸性、穿透性等特點(diǎn)[16],但其獲取的數(shù)據(jù)精度會(huì)受到外界環(huán)境的影響,而且測(cè)量尺度小,在測(cè)量員無(wú)法到達(dá)區(qū)域難以測(cè)量,應(yīng)用具有局限性。無(wú)人機(jī)立體攝影測(cè)量,精度可達(dá)毫米級(jí),但其對(duì)于50~60度陡峭切溝溝壁和懸空的溝壁無(wú)能為力,需要借助地面攝影測(cè)量輔助完成[17],而且還需要在地面設(shè)置很多控制點(diǎn)[18],近年來(lái),隨著無(wú)人機(jī)傾斜攝影測(cè)量技術(shù)逐漸成熟,基于少控制點(diǎn)甚至無(wú)控制點(diǎn)的地形數(shù)據(jù)生成也有報(bào)道,但相較遙感影像而言,其仍然有花費(fèi)時(shí)間長(zhǎng),成本高等缺點(diǎn)。此外,為了保證數(shù)據(jù)的精度,其采集與處理過(guò)程需要在有經(jīng)驗(yàn)的工作人員參與下完成[19]。類似的方法還有近距離移動(dòng)攝影和多視角立體攝影技術(shù)。三維照片重建方法(3D-PR)被用于監(jiān)測(cè)西班牙西南部小型切溝溝頭溯源侵蝕過(guò)程,與地面激光掃描儀對(duì)比驗(yàn)證表明其精度可以達(dá)到厘米級(jí)[20],但是同樣對(duì)地面控制點(diǎn)要求高,而且測(cè)量的空間尺度更小?;谛l(wèi)星遙感立體像對(duì)提取DEM的方法簡(jiǎn)單、快捷且不受區(qū)域限制,同時(shí)也為監(jiān)測(cè)大尺度區(qū)域變化提供了可能性。在國(guó)內(nèi)外已經(jīng)開(kāi)展了基于立體像對(duì)提取DEM精度分析的研究[21-23],主要集中在不同立體像對(duì)提取的DEM精度差異、DEM精度驗(yàn)證方法以及影響立體像對(duì)提取DEM精度的因素等方面[24-27]。目前為止,基于高分衛(wèi)星遙感的立體像對(duì)尚未用于中國(guó)的溝蝕監(jiān)測(cè),提取黃土丘陵區(qū)切溝參數(shù)的精度還不清楚,限制了這項(xiàng)新技術(shù)在黃土高原區(qū)的應(yīng)用。
本文使用半干旱黃土區(qū)小流域高分遙感立體像對(duì)提取切溝的面積、周長(zhǎng)、體積、截面積、溝長(zhǎng)、溝寬、溝深等形態(tài)參數(shù),與三維激光掃描全站儀實(shí)地測(cè)量的數(shù)據(jù)進(jìn)行對(duì)比分析,評(píng)價(jià)高分遙感立體像對(duì)提取的黃土丘陵區(qū)切溝參數(shù)的精度并確定誤差范圍,可為較大尺度上研究黃土高原溝蝕探索實(shí)用技術(shù)和方法。
本文選擇陜北吳起縣退耕還林森林公園內(nèi)的合溝封禁流域作為研究區(qū)(圖1),地理位置為108°12′21″~108°13′55″E,36°53′23″~36°55′27″N,面積約為4.3 km2,海拔在1 290~1 590 m之間。區(qū)域內(nèi)年均氣溫8.0 ℃,1957-2013年均降水量為466.9 mm,降水季節(jié)分布不均,其中7-9月份占全年總降水量的61.6%,屬半干旱溫帶大陸性季風(fēng)氣候,土壤類型以黃綿土為主[13]。研究區(qū)地處黃土高原腹地黃土丘陵第二副區(qū),切溝發(fā)育顯著,溝壑縱橫,屬典型的黃土高原梁狀丘陵溝壑區(qū),溝谷密度7~10 km/km2[28],縣內(nèi)小流域多年平均侵蝕強(qiáng)度為4 399.79 t/(km2·a)[29]。
圖1 研究區(qū)及三維激光掃描區(qū)位置圖
利用托普康IS-IMAGING STATION型三維激光掃描全站儀,于2016年7月對(duì)合溝小流域發(fā)育在溝緣線的21個(gè)切溝進(jìn)行實(shí)地測(cè)量,獲取采樣距離為0.15m的點(diǎn)云數(shù)據(jù)。在Cyclone軟件中對(duì)三維激光掃描儀獲取的每一個(gè)切溝的點(diǎn)云數(shù)據(jù)進(jìn)行去燥處理,去除周圍無(wú)用數(shù)據(jù)以及明顯的誤差點(diǎn)數(shù)據(jù)[30-31]。使用Arcgis10.2將測(cè)量區(qū)域的點(diǎn)云數(shù)據(jù)生成柵格單元為0.15 m的DEM(DEM),用于檢驗(yàn)高分立體像對(duì)提取切溝參數(shù)的精度。為減小植被對(duì)三維激光掃描全站儀測(cè)量結(jié)果的影響,選擇合溝小流域草本覆蓋為主且覆蓋度低的區(qū)域進(jìn)行測(cè)量(圖1)。本次研究所有圖層數(shù)據(jù)均采用高斯-克呂格投影和WGS_1984_UTM_Zone_49N坐標(biāo)系。
使用GeoEye-1立體像對(duì),空間分辨率為0.5 m,拍攝于2016年3月10日。3月份的合溝小流域植被覆蓋率低,降低了植被對(duì)提取的DEM精度的影響。用Erdas9.2對(duì)立體像對(duì)的左右兩片進(jìn)行裁剪后,再利用ENVI5.1的DEM Extraction模塊分別輸入控制點(diǎn)與連接點(diǎn)后自動(dòng)提取分辨率為0.5 m的DEM,本次研究輸入了4個(gè)控制點(diǎn),分別位于三維激光掃描儀測(cè)量區(qū)域的中間位置(圖1),60個(gè)連接點(diǎn),連接點(diǎn)誤差值最大值小于0.5 m。
為了減小切溝邊緣的誤差,在分辨率為0.5 m的DEM基礎(chǔ)上,使用ArcGIS的Resample模塊,選擇最鄰近法對(duì)立體像對(duì)DEM重采樣生成與三維激光掃描數(shù)據(jù)一致的分辨率為0.15 m的DEM。最鄰近法重采樣適用于離散和連續(xù)值數(shù)據(jù)且不會(huì)改變柵格數(shù)據(jù)集中的數(shù)值,對(duì)于立體像對(duì)提取DEM,采用最鄰近法重采樣更為適合[32]。將分辨率為0.5 m的DEM,重采樣為0.15 m,不會(huì)降低DEM的精度,但可以提高地形劇烈變化的切溝邊界等地貌部位的柵格密度,使其與三維激光掃描數(shù)據(jù)生成的DEM的柵格密度一致,從而降低提取切溝邊界以及裁剪切溝DEM的誤差。
分別從三維激光掃描數(shù)據(jù)和高分立體遙感像對(duì)提取切溝參數(shù)的步驟如下。首先,為了保證三維激光掃描數(shù)據(jù)生成的DEM與高分立體像對(duì)生成的DEM精確配準(zhǔn),從高分立體遙感影像獲取的每個(gè)切溝溝頭和切溝兩側(cè)分別選取1個(gè)控制點(diǎn)(共3個(gè)),使用ArcGIS的Georeferencing模塊分別對(duì)三維激光掃描儀獲取的21條切溝進(jìn)行配準(zhǔn)處理。然后,根據(jù)李鎮(zhèn)等[16]確定切溝邊界的方法,從三維激光掃描數(shù)據(jù)生成的DEM上目視解譯并勾繪21條切溝二維邊界,同時(shí),結(jié)合GeoEye-1影像,在高分立體像對(duì)生成的DEM上勾繪出相對(duì)應(yīng)的21條切溝的二維邊界,計(jì)算切溝的二維參數(shù)(面積、周長(zhǎng)、溝寬和溝長(zhǎng)等)。最后,按照如下方法計(jì)算切溝三維參數(shù):利用切溝邊界分別裁剪三維激光掃描數(shù)據(jù)生成的DEM和高分立體像對(duì)生成的DEM,分別得到切溝DEM(DEM),將切溝邊界轉(zhuǎn)換為點(diǎn)圖層并根據(jù)立體影像生成的DEM給點(diǎn)圖層賦予高程值,基于切溝邊界高程點(diǎn)數(shù)據(jù)構(gòu)建大小為0.15m的DEM,再利用切溝邊界圖層裁剪得到切溝原始侵蝕基準(zhǔn)面(DEM0),計(jì)算DEM與DEM0的體積VDEM與VDEM0,切溝體積=VDEM-VDEM0。使用ArcScene建立切溝三維模型,根據(jù)ArcScene提供的測(cè)量工具測(cè)量最大溝深;在每個(gè)切溝溝頭最大寬度部位和切溝中部提取2個(gè)橫截面,分別測(cè)量溝頂寬、溝底寬,平均溝深和橫截面積。通過(guò)與DEM提取的參數(shù)進(jìn)行對(duì)比分析,檢驗(yàn)使用立體像對(duì)提取黃土區(qū)切溝參數(shù)的可靠性。此外,選擇A(體積誤差最大),B(體積誤差中等)和C(體積誤差最?。?個(gè)切溝,結(jié)合ArcScene三維顯示模塊,分析造成誤差的主要原因。除了最常用的誤差以外,還選擇3個(gè)誤差評(píng)價(jià)指標(biāo),公式如下[33]。
式中RMSE表示標(biāo)準(zhǔn)誤差(均方根誤差),mean表示平均誤差;表示百分誤差;Z表示立體像對(duì)DEM中提取的第條切溝的形態(tài)參數(shù);z表示DEM中提取的相對(duì)應(yīng)的第條切溝的形態(tài)參數(shù);表示切溝數(shù)(=21)。
與三維激光測(cè)量相比,基于高分衛(wèi)星遙感立體像對(duì)測(cè)量切溝面積的平均誤差為3.58 m2,周長(zhǎng)的平均誤差為0.55 m,溝長(zhǎng)和溝頂寬的平均誤差分別為0.13和?0.10 m(表1)。
表1 基于立體像對(duì)測(cè)量切溝形態(tài)參數(shù)的誤差統(tǒng)計(jì)
注:* 統(tǒng)計(jì)樣本數(shù)=21。
Note: Statistic sample=21.
切溝線狀和面狀形態(tài)參數(shù)百分誤差()分級(jí)統(tǒng)計(jì)結(jié)果如圖2所示。
圖2 基于高分遙感立體像對(duì)提取的切溝線狀和面狀參數(shù)百分誤差頻率分布
由圖2可知,切溝周長(zhǎng)和溝長(zhǎng)的誤差最低,周長(zhǎng)范圍為0.41%~18.58%,溝長(zhǎng)范圍為0.10%~12.28%,平均百分誤差都小于5%,周長(zhǎng)和溝長(zhǎng)δ主要分布在5%以內(nèi),分別占70.83%和83.33%。相對(duì)于周長(zhǎng)和溝長(zhǎng)來(lái)說(shuō),切溝面積和溝頂寬的誤差較大,面積范圍為0.11%~33.81%,有58.33%的樣本百分誤差在5%以內(nèi)。溝頂寬δ范圍為0.09%~37.41%,溝頂寬主要分布在10%之間,所占樣本總數(shù)的61.9%。從總體上看,切溝線狀和面狀參數(shù)百分誤差可以維持在較低的水平。
與三維激光全站儀獲得的參數(shù)相比,高分衛(wèi)星遙感立體像對(duì)提取的切溝溝深、截面積、體積等切溝三維形態(tài)參數(shù)百分誤差如圖3所示。最大溝深平均誤差為0.14 m,但平均溝深的平均誤差較大,為?0.46 m,二者的百分誤差主要分布在30%以內(nèi),分別占57.14%和54.76%。由表1可知,溝底寬誤差整體大于溝頂寬,平均誤差為?0.67 m,表明立體像對(duì)提取切溝溝底形態(tài)誤差較大,同時(shí)提取切溝的溝深偏小。相應(yīng)地,溝頂寬與溝底寬比率、寬深比、橫截面積和切溝體積的誤差都比較大(圖3)。
切溝寬深比是表征切溝形態(tài)的主要參數(shù),可以用于判斷切溝發(fā)育的階段[2]。根據(jù)三維激光測(cè)量的結(jié)果,21條切溝的寬深比介于1.43~3.83,但立體像對(duì)分析結(jié)果介于1.03~5.32,寬深比百分誤差的平均值為38.26%,最大百分誤差達(dá)到167.8%(圖3)。溝頂寬與溝底寬比率用于判斷切溝橫斷面為“V”型還是“U”型。根據(jù)激光測(cè)量的結(jié)果,溝頂寬與溝底寬比率介于1.18~7.04,立體像對(duì)測(cè)量結(jié)果介于1.16~7.45,但是單個(gè)切溝測(cè)量的百分誤差均值達(dá)到57.13%,最大誤差達(dá)到534%(圖3)。橫截面積的平均誤差為?6.30 m2,百分誤差均值為37.96%。體積的平均誤差為?54.01%,百分誤差介于1.98~88.72%,均值為37.46%。截面積和體積百分誤差主要還能控制在50%以內(nèi),分別占樣本總數(shù)的69.05%和75.00%??偟膩?lái)看,高分遙感立體像對(duì)提取三維切溝參數(shù)的誤差較大。
圖3 基于立體像對(duì)測(cè)量三維切溝參數(shù)的百分誤差頻率分布
立體像對(duì)測(cè)量切溝三維參數(shù)的誤差主要源于立體像對(duì)提取切溝溝底形態(tài)誤差較大,溝深和溝底寬測(cè)量值明顯偏小,其中,在42個(gè)橫截面中,66%的平均溝深小于三維激光掃描測(cè)量值。相應(yīng)地,切溝體積和截面積的測(cè)量值都偏小。圖4中GA的立體像對(duì)測(cè)量切溝體積的誤差是?111.4 m3,百分誤差為49.32%,主要是因?yàn)闇项^上部溝底形態(tài)測(cè)量誤差較大,平均溝深和溝底寬都小于三維激光測(cè)量值;圖4中GB的立體像對(duì)測(cè)量切溝體積的誤差是?16.4 m3,體積百分誤差為23.26%。最大溝深的誤差1.29 m,但溝頭上部的平均溝深明顯小于三維激光測(cè)量值,溝底形態(tài)和橫截面積的誤差都較大;圖4中GC立體像對(duì)測(cè)量切溝體積的誤差是?12.9 m3,體積百分誤差為12.07%,溝底和橫截面積的誤差相對(duì)較小。
注:GAdem,GBdem,GCdem分別指三維激光掃描儀提取A、B和C的DEM二維顯示圖;GA3dscan,GB3dscan,GC3dscan指三維激光掃描儀提取切溝的三維模型圖;GAssi,GBssi,GC3ssi指遙感立體像對(duì)提取切溝三維模型圖.
雖然高分衛(wèi)星遙感立體像對(duì)測(cè)量三維切溝形態(tài)參數(shù)的誤差較大,但進(jìn)一步分析表明切溝體積、截面積和溝底寬的測(cè)量誤差都與切溝規(guī)模關(guān)系比較密切(圖5),在0.05的水平上,具有顯著的線性相關(guān)性。切溝規(guī)模越大,切溝體積、截面積和溝底寬的測(cè)量值偏小的幅度越大。尤其是切溝體積測(cè)量誤差與切溝體積之間可以建立較好的線性回歸模型,在精度要求不高且缺少其他測(cè)量手段時(shí),可以使用該模型對(duì)測(cè)量誤差進(jìn)行校正。
圖5 切溝三維形態(tài)參數(shù)與誤差的關(guān)系
高分遙感立體像對(duì)應(yīng)用于地形地貌測(cè)量方面已取得較大進(jìn)展。Haque等[34]利用分辨率為0.5 m的GeoEye-1立體像對(duì)提取孟加拉丘陵區(qū)DEM并以GPS實(shí)測(cè)點(diǎn)作參照,確定DEM精度為1.25m;Fiorucci等[35]利用WorldView-1和GeoEye-1立體像對(duì)提取意大利中部河谷丘陵區(qū)的DEM,通過(guò)與地面控制點(diǎn)作對(duì)比分析得到DEM誤差為2.01 m,在此基礎(chǔ)上測(cè)量了小流域尺度的細(xì)溝和淺溝侵蝕量。相對(duì)于上述地區(qū),黃土丘陵區(qū)地形更加破碎,應(yīng)用立體像對(duì)提取地形地貌參數(shù)的研究還很少。本文采用分辨率為0.5 m的高分GeoEye-1立體像對(duì)提取黃土丘陵區(qū)的合溝小流域DEM,并與三維激光掃描儀實(shí)測(cè)數(shù)據(jù)獲得的精度為0.15 m的DEM作參照,分析利用立體像對(duì)提取黃土丘陵區(qū)切溝形態(tài)參數(shù)的精度。其中,線狀參數(shù)的誤差介于0.13 m(溝寬平均誤差)和0.55 m(切溝周長(zhǎng)平均誤差)之間。與上述兩項(xiàng)研究結(jié)果有一定可比性[34-35]。高分遙感立體像對(duì)測(cè)量切溝長(zhǎng)度、周長(zhǎng)和面積的百分誤差主要分布在5%以內(nèi),可以用于黃土區(qū)切溝線狀和面狀參數(shù)的測(cè)量??墒牵S切溝形態(tài)參數(shù)的測(cè)量誤差較大,尤其是溝底寬、寬深比、橫截面積的平均百分誤差在30%左右,難以保證切溝三維參數(shù)的測(cè)量精度。
在切溝參數(shù)的實(shí)際測(cè)量中有多種影響測(cè)量精度的因素。首先,使用的遙感立體像對(duì)的空間分辨率是0.5 m,無(wú)法精確獲取黃土區(qū)溝谷的微地形變化。其次,由于拍攝角度的問(wèn)題,提取的DEM在某些地貌部位可能存在變形,而正射校正的精度有限,也是影響DEM精度的因素之一。再次,惠鳳鳴等[36]研究表明地面控制點(diǎn)的數(shù)量越多,分布越均勻,提取的DEM精度越高,本研究中由于基礎(chǔ)資料所限,用于立體像對(duì)提取DEM的控制點(diǎn)數(shù)量有限且位置相對(duì)集中,因此會(huì)在一定程度上影響提取的DEM精度。最后,切溝自身形態(tài)對(duì)提取的切溝參數(shù)精度也存在一定的影響。本文基于發(fā)育在溝緣線上的切溝體積測(cè)量誤差與切溝體積之間的線性關(guān)系,建立了針對(duì)高分遙感立體像對(duì)測(cè)量切溝體積的校正模型,但是,不同地區(qū)的切溝形態(tài),同一地區(qū)發(fā)育在不同地貌部位的切溝形態(tài)都有差異,因此,本文提出校正模型以及切溝參數(shù)測(cè)量精度只能適用于黃土丘陵區(qū)發(fā)育在溝緣線的切溝,但可以為不同類型切溝測(cè)量提供參考。另外,與地面三維激光掃描數(shù)據(jù)對(duì)比,分析高分衛(wèi)星遙感立體像對(duì)的測(cè)量精度,是基于已有研究結(jié)果假定地面三維激光掃描能夠準(zhǔn)確測(cè)量切溝參數(shù)[13-16]。但是三維激光掃描儀在采集數(shù)據(jù)時(shí),也會(huì)受到植被、掃描角度等因素的干擾,所以用于檢驗(yàn)的三維地形數(shù)據(jù)本身存在一定的誤差,從而對(duì)精度評(píng)價(jià)結(jié)果造成一定影響。
本文以三維激光掃描全站儀采集的間距為0.15 m三維地形數(shù)據(jù)為對(duì)照,分析0.5 m分辨率的高分遙感立體像對(duì)提取黃土丘陵區(qū)小流域切溝參數(shù)的精度。結(jié)果如下:
1)在高分遙感立體像對(duì)提取的線狀和面狀切溝參數(shù)中,面積、周長(zhǎng)、溝長(zhǎng)和溝頂寬的平均誤差分別為3.58 m2,0.55 m,0.13 m和?0.10 m。面積、周長(zhǎng)和溝長(zhǎng)的百分誤差主要分布在5%以內(nèi),溝頂寬百分誤差主要分布在10%以內(nèi),具有較高精度。
2)三維切溝參數(shù)中,溝底寬、最大溝深、平均溝深的平均誤差分別為?0.67、0.14和?0.46 m。截面積和體積的平均誤差分別為?6.30 m2和?54.01 m3。最大溝深百分誤差主要集中在30%以內(nèi),溝底寬、平均溝深、截面積和體積百分誤差則主要分布在50%以內(nèi)。相較于三維激光掃描切溝,立體像對(duì)測(cè)量切溝形態(tài)的溝底誤差較大,主要是溝底寬和平均溝深偏小。
3)切溝規(guī)模越大,切溝體積、截面積和溝底寬的測(cè)量值偏小的幅度越大。切溝體積測(cè)量誤差與切溝體積之間可以建立較好的線性回歸模型,在缺少其他測(cè)量手段時(shí),可以使用該模型對(duì)測(cè)量誤差進(jìn)行校正。
立體像對(duì)的空間分辨率、正射校正的精度,控制點(diǎn)數(shù)量等因素是造成高分遙感立體測(cè)量產(chǎn)生較大誤差的主要原因。但高分立體遙感為切溝線狀和面狀參數(shù)測(cè)量以及切溝體積測(cè)量提供了新的方法,為黃土丘陵區(qū)溝蝕監(jiān)測(cè)提供了便捷、且相對(duì)可靠的數(shù)據(jù)源。
[1] 景可. 黃土高原溝谷侵蝕研究[J]. 地理科學(xué),1986,6(4):340-347.
Jing Ke. A study on gully erosion on the Loess Plateau[J]. Scientia Geographic Sinica, 1986, 6(4): 340-347. (in Chinese with English abstract)
[2] 朱顯謨. 黃土區(qū)土壤侵蝕的分類[J]. 土壤學(xué)報(bào),1956,4(2):99-115.
Zhu Xianmo. Classification on the soil erosion in the Loess Plateau[J]. Acta Pedologica Sinica, 1956, 4(2): 99-115. (in Chinese with English abstract)
[3] 羅來(lái)興. 劃分晉西、陜北、隴東黃土區(qū)域溝間地與溝谷的地貌類型[J]. 地理學(xué)報(bào),1956,22(3):201-221.
Luo Laixing. A tentative classification of landforms in the Loess Plateau[J]. Acta Geographica Sinica, 1956, 22(3): 201-221. (in Chinese with English abstract)
[4] 張新和,鄭粉莉,李靖. 切溝侵蝕研究現(xiàn)狀與存在問(wèn)題分析[J]. 水土保持研究,2007,14(4):31-32,36.
[5] Martinez-Casasnovas J A, Ramos M C, Poesen J. Assessment of sidewall erosion in large gullies using multi-temporal DEMs and logistic regression analysis[J]. Geomorphology, 2004, 58(1/4): 305-321.
[6] Whitford J A, NewhamL T H, Vigiak O, et al. Rapid assessment of gully sidewall erosion rates in data-poor catchments: A case study in Australia[J]. Geomorphology, 2010, 118(3/4): 330-338.
[7] Poesen J, Nachtergaele J, Verstraeten G, et al. Gully erosion and environmental change: Importance and research needs[J]. Catena, 2003, 50: 91-133.
[8] Castillo C, Gómez J A. A century of gully erosion research: Urgency, complexity and study approaches[J]. Earth-Science Reviews, 2016, 160: 300-319.
[9] 馬保東,吳立新,許志華. 利用資源三號(hào)測(cè)繪衛(wèi)星立體像對(duì)提取DEM及精度評(píng)價(jià):以神東礦區(qū)大柳塔礦為例[J]. 測(cè)繪通報(bào),2013(11):68-70,77.
[10] 湯國(guó)安. 我國(guó)數(shù)字高程模型與數(shù)字地形分析研究進(jìn)展[J]. 地理學(xué)報(bào),2014,69(9):1305-1325.
Tang Guoan. Progress of DEM and digital terrain analysis in China[J]. Acta Geographica Sinica, 2014, 69(9): 1305-1325. (in Chinese with English abstract)
[11] 胡剛,伍永秋,劉寶元,等. GPS和GIS進(jìn)行短期溝蝕研究初探:以東北漫川漫崗黑土區(qū)為例[J]. 水土保持學(xué)報(bào),2004,18(4):16-19.
Hu Gang, Wu Yongqiu, Liu Baoyuan, et al. Preliminary research on short-term channel erosion using GPS and GIS[J]. Journal of Soil and Water Conservation, 2004, 18(4): 16-19. (in Chinese with English abstract)
[12] Wu Yongqiu, Cheng Hong. Monitoring of gully erosion on the Loess Plateau of China using a global positioning system[J]. Catena, 2005, 63: 154-166.
[13] Jackson T J, Ritchie J C, White J, et al. Airborne laser profile data for measuring ephemeral gully erosion[J]. Photogrammetric Engineering and Remote Sensing, 1988, 54(8): 1181-1185.
[14] Romanescu G, Cotiuga V, Asandulesei A, et al. Use of the 3-D scanner in mapping and monitoring the dynamic degradation of soils: Case study of the Cucuteni-Baiceni Gully on the Moldavian Plateau (Romania)[J]. Hydrology and Earth System Sciences, 2012, 16(3): 953-966.
[15] 劉希林,張大林. 基于三維激光掃描的崩崗侵蝕的時(shí)空分析[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(4):204-211.
Liu Xilin, Zhang Dalin. Temporal-spatial analyses of collapsed gully erosion based on three-dimensional laser scanning[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(4): 204-211. (in Chinese with English abstract)
[16] 李鎮(zhèn),張巖,楊松,等. QuickBird影像目視解譯法提取切溝形態(tài)參數(shù)的精度分析[J]. 農(nóng)業(yè)工程學(xué)報(bào),2014,30(20):179-186.
Li Zhen, Zhang Yan, Yang Song, et al. Error assessment of extracting morphological parameters of bank gullies by manual visual interpretation based on QuickBird imagery[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2014, 30(20): 179-186. (in Chinese with English abstract)
[17] St?cker C, Eltner A, Karrasch P. Measuring gullies by synergetic application of UAV and close range photogrammetry: A case study from Andalusia, Spain[J]. Catena, 2015, 132: 1-11.
[18] Kaiser, Andreas, Neugirg, et al. Small-scale surface reconstruction and volume calculation of soil erosion in complex moroccan gully morphology using structure from motion[J]. Remote Sensing, 2014, 6(8): 7050-7080.
[19] Liu Kai, Ding Hui, Tang Guoan, et al. Detection of catchment-scale gully-affected areas using unmanned aerial vehicle (UAV) on the Chinese Loess Plateau[J]. ISPRS International Journal of Geo-Information, 2016, 5(12): 238.
[20] álvaro Gómez-Gutiérrez, Schnabel S, Berenguer-Sempere F, et al. Using 3D photo-reconstruction methods to estimate gully headcut erosion[J]. Catena, 2014, 120(1): 91-101.
[21] Tan C E J, Jafri M Z M, Lim H S, et al. Digital elevation model (DEM) generation from stereo images[J]. Pertanika Journal of Science & Technology, 2011, 19(S): 77-82.
[22] Hu Fen, Gao Xiaoming, Li Guoyuan, et al. DEM extraction from WORLDVIEW-3 stereo-images and accuracy evaluation[J]. International Archives of the Photogrammetry Remote Sensing & S, 2016, XLI-B1: 327-332.
[23] Dong Y, Chen W, Chang H, et al.Assessment of orthoimage and DEM derived from ZY-3 stereo image in Northeastern China[J]. Survey Review, 2016, 48(349): 247-257.
[24] 王廣杰,何政偉,仇文俠,等. ASTER立體像對(duì)提取瑪爾擋壩區(qū)DEM及精度評(píng)價(jià)[J]. 測(cè)繪科學(xué),2009,34(3):94-96.
Wang Guangjie, He Zhengwei, Qiu Wenxia, et al. Extraction and accuracy assessment of DEM on maerdang dam area from ASTER stereo image data[J]. Science of Surveying & Mapping, 2009, 34(3): 94-96. (in Chinese with English abstract)
[25] 賀雪艷,張路,Timo Balz. 利用外部DEM輔助山區(qū)SAR立體像對(duì)匹配及地形制圖[J]. 測(cè)繪學(xué)報(bào),2013,42(3):425-432.
He Xueyan, Zhang Lu, Balz Timo. Topographic mapping in mountainous areas using stereoSAR assisted by external DEM[J]. Acta Geodaetica et Cartographica Sinica, 2013, 42(3): 425-432. (in Chinese with English abstract)
[26] 夏濤,楊武年,劉漢湖,等. 利用ASTER立體像對(duì)提取相對(duì)DEM及正射影像地圖制作[J]. 測(cè)繪科學(xué),2007,32(3):144-145.
Xia Tao, Yang Wunian, Liu Hanhu, et al. Automatic digital elevation model (DEM) and orthoimage generation from ASTER image[J]. Science of Surveying & Mapping, 2007, 32(3): 144-145. (in Chinese with English abstract)
[27] 丁輝,姚安強(qiáng). 利用IKONOS立體像對(duì)提取DEM精度的實(shí)驗(yàn)[J]. 測(cè)繪科學(xué),2012(1):179-181.
Ding Hui, Yao Anqiang. DEM generation and analysis using IKONOS stereo pairs[J]. Science of Surveying and Mapping, 2012(1): 179-181. (in Chinese with English abstract)
[28] 田劍,湯國(guó)安,周毅,等. 黃土高原溝谷密度空間分異特征研究[J]. 地理科學(xué),2013,33(5):622-628.
Tian Jian, Tang Guoan, Zhou Yi, et al. Spatial variation of gully density in the loess plateau[J]. Scientia Geographica Sinica, 2013, 33(5): 622-628. (in Chinese with English abstract)
[29] 秦偉,朱清科,張巖. 基于GIS和RUSLE的黃土高原小流域土壤侵蝕評(píng)估[J]. 農(nóng)業(yè)工程學(xué)報(bào),2009,25(8):157-163. Qin Wei, Zhu Qingke, Zhang Yan. Soil erosion assessment of small watershed in Loess Plateau based on GIS and RUSLE[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2009, 25(8): 157-163. (in Chinese with English abstract)
[30] 張鵬,鄭粉莉,王彬,等. 高精度GPS三維激光掃描和測(cè)針板三種測(cè)量技術(shù)監(jiān)測(cè)溝蝕過(guò)程的對(duì)比研究[J]. 水土保持通報(bào),2008,28(5):11-15.
Zhang Peng, Zheng Fenli, Wang Bin, et al. Comparative study of monitoring gully erosion morphology change process by using high precision GPS, Leica HDS 3000 laser scanner and needle board method[J]. Bulletin of Soil & Water Conservation, 2008, 28(5): 11-15. (in Chinese with English abstract)
[31] 陳俊杰,孫莉英,劉俊體,等. 坡度對(duì)坡面細(xì)溝侵蝕的影響:基于三維激光掃描技術(shù)[J]. 中國(guó)水土保持科學(xué),2013,11(3):1-5.
Chen Junjie, Sun Liying, Liu Junti, et al. Effects of slope gradients on rill erosion: Study based on three-dimensional laser technology[J]. Science of Soil & Water Conservation, 2013, 11(3): 1-5. (in Chinese with English abstract)
[32] 鄒曉軍. 攝影測(cè)量與遙感[M]. 北京:測(cè)繪出版社,2011.
[33] Haberlandt U. Geostatistical interpolation of hourly precipitation from rain gauges and radar for a large-scale extreme rainfall event[J]. Journal of Hydrology, 2007, 332(1/2): 144-157.
[34] Haque M F, Alam M S, Quadir D A. Accuracy assessment of generated DEMs and ortho-images from GeoEye-1 stereo pair for a selected hilly area of Bangladesh[J]. Science, 2014, 3: 2277-8179.
[35] Fiorucci F, Ardizzone F, Rossi M, et al. The use of stereoscopic satellite images to map rills and ephemeral gullies[J]. Remote Sensing, 2015, 7(10): 14151-14178.
[36] 惠鳳鳴,田慶久,李應(yīng)成. Aster數(shù)據(jù)的DEM生產(chǎn)及精度評(píng)價(jià)[J]. 遙感信息,2004(1):14-18.
Hui Fengming, Tian Qingjiu, Li Yingcheng. Production and accuracy assessment of DEM from ASTER stereo image data[J]. Remote Sensing Information, 2004(1): 14-18. (in Chinese with English abstract)
Accuracy assessment of gully morphological parameters from high resolution remote sensing stereoscopic satellite images on hilly Loess Plateau
Tang Jie, Zhang Yan※, Fan Conghui, Cheng Xiaoxin, Deng Jiayong
(,,100083)
As one of the main forms of soil erosion on the hilly Loess Plateau, gully erosion threatens the land resources and causes serious environmental problems. Although different approaches have been used to monitor and predict gully erosion over the years, gully erosion measurement on middle and large scale is still difficult. The objective of this study was to assess the accuracy of measuring gully morphological parameters using stereoscopic satellite images with very high resolution on the hilly Loess Plateau and feasibility of stereoscopic satellite images acting a reliable data source for large scale gully erosion monitoring. GeoEye-1 stereoscopic satellite images (0.5 m panchromatic and 2.0 m multispectral images, acquired in March, 2016) were used to calculate gully morphological parameters of 21 gullies in a catchment located in central Loess Plateau and then test their accuracy by comparing them with 3D (three-dimensional) laser scanning data (0.15 m resolution, acquired in July, 2016), and mean error, percent error and root mean squared error were used for error measure. The results were as follows: 1) The average errors of linear and areal gully parameters, including area, perimeter, length and width, were 3.58 m2, 0.55 m, 0.13 m and -0.10 m, respectively, and the percentage errors were 0.11%-33.81%, 0.41%-18.58%, 0.10%-12.28%, and 0.09%-37.41%, with the average percentage error of 8.96%, 4.83%, 2.92%, and 10.81%, respectively. The percentage errors of area, perimeter, and length were mainly below 5%, and the percentage errors of gully width were mainly below 10%. Overall, the errors of linear and areal gully parameters extracted from high resolution stereoscopic satellite images can be controlled in a lower level. 2) The average errors of 3D gully parameters, including gully bottom width, maximum gully depth, average gully depth, cross sectional area and gully volume, were -0.67 m, 0.14 m, -0.46 m, -6.30 m2and -54.01 m3, respectively. The percentage errors of 3D gully parameters were 0.39%-84.65%, 0.51%-55.91%, 0-59.64%, 0.13%-81.53%, and 1.98%-88.72%, respectively, with the average percentage error of 30.37%, 24.13%, 27.16%, 37.96% and 37.46%, respectively. The percentage errors of 3D gully parameters were mainly below 50% except the maximum gully depth which was mainly below 30%. Compared with 3D laser scanning, gully bottoms morphology measured with stereoscopic satellite images presented smaller gully depths, gully bottom width and cross sectional area. 3) 3D gully parameters including gully volume, cross sectional area and gully bottom width were correlated significantly with gully scale at the level of 0.05. The larger the scale of the gullies, the smaller the measured gully volume, cross sectional area and gully bottom width from stereoscopic satellite images than 3D laser scanning measurement. A linear regression model was built between measurement errors and gully volumes, which can be used when other measurement approaches are not available. On the whole, the accuracy of measured gully parameters from stereoscopic satellite images will be affected by many factors, such as the resolution of stereoscopic satellite images, the correction accuracy and the number of control points, but it also can provide reliable data for linear and areal gully parameters measurement and new method for 3D parameters measurement of large scale gullies on the hilly Loess Plateau.
remote sensing; measurements; abstracts; 3D laser scanning; gully morphological parameters; measurement errors; hilly Loess Plateau
10.11975/j.issn.1002-6819.2017.18.015
S157
A
1002-6819(2017)-18-0111-07
2017-05-04
2016-09-13
國(guó)家自然科學(xué)基金項(xiàng)目(41671272);國(guó)家重點(diǎn)研發(fā)計(jì)劃(2016YFC0501604-05)
唐 杰,主要研究方向?yàn)橘Y源環(huán)境規(guī)劃與管理。Email:jayetang@qq.com
張 巖,博士,教授,主要研究方向?yàn)橥寥狼治g和水土保持。Email:zhangyan9@bjfu.edu.cn