黨旭偉 林馨園 賀正 陳燕 慈寶霞 馬學(xué)花 郭晨荔 賀亞星 劉揚(yáng) 馬富裕
doi:10.6048/j.issn.1001-4330.2024.03.005
摘? 要:【目的】提高基于熱紅外遙感圖像滴灌棉花冠層溫度提取精度,為棉花水分狀況精準(zhǔn)監(jiān)測(cè)提供技術(shù)支撐。
【方法】以不同水分處理的苗期、蕾期棉花為研究對(duì)象,利用無人機(jī)獲取試驗(yàn)小區(qū)熱紅外遙感圖像,使用便攜式手持測(cè)溫儀測(cè)量田間輻射校正板及水桶中的水溫,對(duì)熱紅外影像進(jìn)行溫度校正。采用Otsu算法、Canny邊緣檢測(cè)算法對(duì)熱紅外遙感圖像進(jìn)行掩膜處理剔除土壤背景,通過波段運(yùn)算提取棉花冠層溫度,繪制棉花冠層溫度頻率直方圖并對(duì)其進(jìn)行優(yōu)化。利用便攜式手持測(cè)溫儀同步測(cè)量棉花冠層溫度,與提取的冠層溫度進(jìn)行一致性分析,驗(yàn)證熱紅外遙感圖像提取棉花冠層溫度的精度。
【結(jié)果】Canny邊緣檢測(cè)算法剔除土壤背景提取冠層圖像準(zhǔn)確率大于Otsu算法(91.90%>82.52%、92.76%>80.60%),剔除土壤背景效果最優(yōu)。Otsu算法和Canny邊緣檢測(cè)算法剔除土壤背景后構(gòu)建的冠層溫度直方圖均呈偏態(tài)分布,但Canny邊緣檢測(cè)算法剔除土壤背景后構(gòu)建的冠層溫度直方圖形狀比Otsu算法光滑,噪聲少,并且Canny邊緣檢測(cè)算法2年冠層平均溫度最低(29.95、30.54℃),與實(shí)測(cè)溫度差值最?。?.78、3.43℃)。去除Canny邊緣檢測(cè)算法的溫度直方圖兩端1%溫度信息后,提取的冠層溫度與實(shí)測(cè)溫度相關(guān)性最高(2年試驗(yàn)r由0.88、0.93提高到了0.94、0.95),RMSE最低(2年RMSE由2.78、2.87℃下降到1.59、1.43℃)。
【結(jié)論】Canny邊緣檢測(cè)算法提高了無人機(jī)熱紅外遙感圖像棉花冠層溫度提取精度,且溫度直方圖兩端1%溫度優(yōu)化后有助于提高棉花冠層溫度提取精度。
關(guān)鍵詞:滴灌棉花;無人機(jī);熱紅外遙感圖像;冠層溫度;精度
中圖分類號(hào):S562;S123??? 文獻(xiàn)標(biāo)志碼:A??? 文章編號(hào):1001-4330(2024)03-0565-11
收稿日期(Received):
2023-07-15
基金項(xiàng)目:
新疆生產(chǎn)建設(shè)兵團(tuán)財(cái)政科技計(jì)劃項(xiàng)目“新疆高品質(zhì)棉花生產(chǎn)關(guān)鍵栽培技術(shù)集成與示范”(2020AB017);“主要作物精準(zhǔn)水肥一體化技術(shù)和裝備研發(fā)及應(yīng)用示范”(2020AB018)
作者簡(jiǎn)介:
黨旭偉(1997-),男,新疆石河子人,碩士研究生,研究方向?yàn)樽魑锼指咝Ю门c精準(zhǔn)栽培,(E-mail)dxw202009@163.com
通訊作者:
劉揚(yáng)(1989-),女,青海民和人,副教授,博士,碩士生導(dǎo)師,研究方向?yàn)樽魑锼矢咝Ю眉熬珳?zhǔn)栽培,(E-mail)ly.0318@163.com
馬富裕(1967-),男,甘肅環(huán)縣人,教授,博士,碩士生/博士生導(dǎo)師,研究方向?yàn)樽魑锞珳?zhǔn)栽培與水肥一體化,(E-mail)1469633844@qq.com
0? 引 言
【研究意義】快速、準(zhǔn)確的獲取冠層溫度,可以判斷作物水分狀況,指導(dǎo)農(nóng)田灌溉[1-4]。傳統(tǒng)冠層溫度主要使用手持式紅外測(cè)溫儀獲取,但僅局限于點(diǎn)測(cè)溫,無法評(píng)估整塊田地冠層溫度分布情況,冠層溫度信息代表性較差[5,6]。通過無人機(jī)搭載熱紅外傳感器可快速、及時(shí)、大范圍獲取作物冠層熱紅外遙感圖像,并用于作物水分狀況精準(zhǔn)監(jiān)測(cè)。然而熱紅外遙感圖像中含有大量的土壤背景、作物冠層等地物信息,導(dǎo)致提取的冠層溫度信息中存在地物混合像元,對(duì)冠層溫度的提取形成干擾,因此剔除無人機(jī)熱紅外遙感圖像中的土壤背景,對(duì)提高作物冠層溫度提取精度至關(guān)重要[7-9]?!厩叭搜芯窟M(jìn)展】目前,剔除無人機(jī)熱紅外遙感圖像土壤背景常采用3種方法:一是感興趣區(qū)域選擇,例如在提取作物冠層溫度時(shí),通過ROI來手動(dòng)劃分熱紅外遙感圖像中葉片覆蓋區(qū)域,進(jìn)而計(jì)算該區(qū)域內(nèi)的平均溫度,即為作物冠層溫度,雖然ROI可以消除部分土壤背景的干擾,但針對(duì)葉片縫隙之間包含的土壤背景仍然無法剔除[10-13];二是基于可見光圖像降噪的冠層溫度掩膜,采用RGRI指數(shù)法、RGBI指數(shù)法、EXG算法剔除作物可見光圖像土壤背景,之后將熱紅外遙感圖像冠層溫度信息掩膜至剔除土壤背景的可見光圖像,可以獲取較高的冠層溫度提取精度,但在預(yù)處理過程中,可見光圖像與熱紅外遙感圖像分辨率不一致,需要降低可見光圖像的分辨率,且需布置較多的控制點(diǎn)對(duì)熱紅外遙感圖像進(jìn)行配準(zhǔn),一旦配準(zhǔn)過程產(chǎn)生偏差,將會(huì)使作物冠層溫度中涵蓋土壤溫度信息 [14-15];三是依據(jù)熱紅外遙感圖像土壤與冠層的灰度差異,直接對(duì)熱紅外遙感圖像降噪進(jìn)而提取冠層溫度,如在棉花花鈴期,利用Otus算法、Canny邊緣檢測(cè)算法等達(dá)到分離土壤與冠層的目的,但是該方法要求圖像分辨率高,冠層與土壤差異明顯,地物信息簡(jiǎn)單[16-17]?!颈狙芯壳腥朦c(diǎn)】由于棉花不同生育時(shí)期冠層結(jié)構(gòu)(冠層開度、葉面積指數(shù)、葉傾角)具有顯著差異,使得土壤背景像元數(shù)量發(fā)生改變,進(jìn)而對(duì)冠層溫度的提取產(chǎn)生干擾。有關(guān)棉花冠層溫度的提取方法研究主要集中于花鈴期,而針對(duì)苗期、蕾期冠層溫度的提取方法研究較少。需突破棉花苗期、蕾期冠層溫度監(jiān)測(cè)方法與技術(shù)研究,實(shí)現(xiàn)早期水分精準(zhǔn)管理,以保證棉花花鈴期、盛鈴期保鈴成鈴高產(chǎn)?!緮M解決的關(guān)鍵問題】以棉花苗期、蕾期冠層相關(guān)數(shù)據(jù)為研究對(duì)象,利用無人機(jī)采集試驗(yàn)小區(qū)熱紅外遙感圖像,分析Otsu算法和Canny邊緣檢測(cè)算法剔除熱紅外遙感圖像土壤背景對(duì)滴灌棉花冠層溫度提取精度的影響,并評(píng)價(jià)冠層溫度提取精度,分析提取冠層溫度最優(yōu)方法,為新疆滴灌棉花水分管理過程中冠層溫度提取和監(jiān)測(cè)提供技術(shù)和數(shù)據(jù)支撐。
1? 材料與方法
1.1? 材 料
1.1.1? 研究區(qū)概況
試驗(yàn)于2021~2022年在石河子大學(xué)教學(xué)試驗(yàn)場(chǎng)(85.97E,44.32N)進(jìn)行,該地區(qū)海拔470 m,屬于典型的溫帶大陸性氣候,冬寒夏熱。年降水量為125~207 mm,平均氣溫25.1~26.1℃,年蒸發(fā)量為1 000~1 500 mm,無霜期為168~171 d,≥0℃的活動(dòng)積溫為4 023~4 118℃,≥10℃的活動(dòng)積溫為3 570~3 729℃。土壤質(zhì)地為黏壤土,平均田間持水量為16.07%(質(zhì)量含水率),平均土壤干容重為1.43 g/cm3。棉花品種為中棉109,2021年4月23日播種5月5日出苗,2022年4月13日播種,4月26日出苗,棉花種植模式為1膜3行,76 cm等行距,播幅2.28 m,全生育期施用 300 kg/hm2N、 108 kg/hm2P2O5、97 kg/hm2K2O ,均隨水滴施,其他管理措施均按照當(dāng)?shù)卮筇锕芾矸绞竭M(jìn)行。圖1,圖2
1.1.2? 無人機(jī)遙感影像及地面數(shù)據(jù)采集
在棉花蕾期(2021年6月3~18日,2022年5月28日~6月15日)于晴朗無風(fēng)日期,采用干濕參考面法,在每個(gè)小區(qū)中間位置選擇4行、行長(zhǎng)1 m的棉花區(qū)域,將其中2行棉花所有葉片正反面均勻涂抹白凡士林作為干參考面,另外2行使用噴壺對(duì)棉花所有葉片噴施清水作為濕參考面,1 min后進(jìn)行無人機(jī)熱紅外遙感圖像和地面數(shù)據(jù)采集,采集時(shí)間為每天12:00~13:00,此外在田間布置有黑白溫度校正板以及測(cè)量水溫的水桶,用于無人機(jī)熱紅外遙感圖像溫度轉(zhuǎn)換與校正。
1.1.2.1? 熱紅外遙感圖像采集
使用大疆御2行業(yè)進(jìn)階版無人機(jī),最大起飛重量1 100 g,續(xù)航時(shí)間可達(dá)31 min。熱紅外相機(jī)工作波段為8~14 μm,像素為640×512,鏡頭焦距38 mm。設(shè)置無人機(jī)飛行高度為30 m,鏡頭垂直地面進(jìn)行拍攝,航向重疊率和旁向重疊率為80%,按照航線規(guī)劃對(duì)試驗(yàn)地進(jìn)行影像采集,飛行時(shí)間在12:30~13:00,采集整個(gè)試驗(yàn)區(qū)域大約用時(shí)15 min。
1.1.2.2? 地面數(shù)據(jù)采集
在各小區(qū)非干濕參考面區(qū)域、干參考面區(qū)域和濕參考面區(qū)域分別選取具有代表性的5株棉花進(jìn)行標(biāo)記,并在無人機(jī)影像采集結(jié)束后,立即使用便攜式手持測(cè)溫儀測(cè)量植株冠層溫度;測(cè)量黑白溫度校正板及水的溫度,用于溫度轉(zhuǎn)換及校準(zhǔn)。
1.2? 方 法
1.2.1? 試驗(yàn)設(shè)計(jì)
設(shè)置4個(gè)水分處理,分別為田間土壤含水量達(dá)80%FC(I1)、田間土壤含水量達(dá)70%FC(I2)、田間土壤含水量達(dá)60%FC(I3)、田間土壤含水量達(dá)50%FC(I4),其中I3為對(duì)照(CK)。每個(gè)處理重復(fù)3次,共計(jì)12個(gè)小區(qū),小區(qū)為6 m×8 m,2個(gè)小區(qū)之間設(shè)有防滲帶。
1.2.2 ?無人機(jī)遙感影像處理
1.2.2.1? 圖像處理
使用Pix 4D mapper軟件對(duì)熱紅外遙感圖像進(jìn)行拼接處理以獲取試驗(yàn)小區(qū)正射影像,并保存為JPEG格式。根據(jù)干濕參考面設(shè)置,將2年拼接成的正射影像通過ENVI軟件裁剪成60個(gè)感興趣區(qū)域。
1.2.2.2? 溫度轉(zhuǎn)換與校準(zhǔn)
使用DJI Thermal Analysis Tool軟件提取熱紅外遙感圖像中的黑白溫度校正板及水的溫度,采用ENVI軟件提取相應(yīng)熱紅外遙感圖像中的黑白溫度校正板及水的灰度,構(gòu)建灰度與溫度的函數(shù)公式,在ENVI軟件中通過波段運(yùn)算將熱紅外遙感圖像由灰度轉(zhuǎn)換為溫度。利用便攜式手持測(cè)溫儀測(cè)量黑白板及水的溫度,每個(gè)物體測(cè)量6次取平均值,與通過DJI Thermal Analysis Tool軟件提取的黑白溫度校正板及水的溫度建立實(shí)測(cè)溫度與圖像溫度的校準(zhǔn)函數(shù),同樣在ENVI軟件中通過波段運(yùn)算完成溫度校正[18]。圖3
1.2.2.3? 剔除土壤背景
Otsu算法:根據(jù)影像的灰度特性,將圖像分割為冠層與土壤兩類,選擇使其類間方差最大、類內(nèi)方差最小的分割閾值為最優(yōu)閾值,進(jìn)行圖像的自動(dòng)二值分類,其核心公式參考程麗娜等[19]。
Canny邊緣檢測(cè)算法:Canny邊緣檢測(cè)算法是一個(gè)多級(jí)邊緣檢測(cè)算法,可依據(jù)作物冠層邊緣特征將圖像分割為冠層與土壤兩類,剔除土壤背景過程主要包括四個(gè)步驟:圖像降噪、計(jì)算圖像梯度向量與梯度幅值、極大值的選擇與非極大值抑制、雙閾值篩選[20,21]。
1.2.2.4? 冠層溫度優(yōu)化
通過Otsu算法和Canny邊緣檢測(cè)算法剔除土壤背景時(shí),會(huì)將一部分土壤背景劃分為棉花冠層,對(duì)溫度提取造成誤差。陰影面土壤溫度低于冠層溫度,陽光直射面土壤溫度高于冠層溫度,且分布于溫度直方圖兩端,通過剔除溫度直方圖兩端溫度信息后,可以提高冠層溫度提取精度。采用剔除溫度直方圖兩端1%冠層溫度直方圖進(jìn)行優(yōu)化,以進(jìn)一步提高冠層溫度提取精度[22]。
1.2.3? 圖像分類精度評(píng)價(jià)
使用photoshop對(duì)棉花冠層熱紅外遙感圖像進(jìn)行人工分割,并將其與算法分割結(jié)果進(jìn)行比對(duì),獲取TP(正確識(shí)別為冠層像元的數(shù)量),F(xiàn)P(將土壤像元識(shí)別為冠層像元的數(shù)量),F(xiàn)n(將冠層像元識(shí)別為土壤像元的數(shù)量),利用混淆矩陣(Confusion Matrix)計(jì)算準(zhǔn)確率(Precision)和召回率(Recall)對(duì)熱紅外遙感圖像冠層像元提取結(jié)果進(jìn)行量化評(píng)價(jià),兩者越接近100%,冠層像元提取精度越高。
Precision=TPTP+FP.(1)
Recall=TPTP+Fn.(2)
1.3? 數(shù)據(jù)處理
通過相關(guān)系數(shù)(r)評(píng)價(jià)兩者的相關(guān)關(guān)系,r越接近1 兩者相關(guān)性越高;通過均方根誤差(Root Mean Square Error,RMSE)評(píng)價(jià)觀測(cè)值的誤差,RMSE越接近0,觀測(cè)值的誤差越小。
2? 結(jié)果與分析
2.1? 冠層區(qū)域提取
2.1.1? 不同算法對(duì)冠層區(qū)域提取方法比較
研究表明,使用Otsu算法剔除土壤背景時(shí),與冠層像元灰度相近的土壤灰度像元被劃分至冠層區(qū)域,造成提取的冠層溫度中既包括陽光直射土壤像元也包括陰影土壤像元,對(duì)冠層溫度提取產(chǎn)生干擾。Canny算法可以精準(zhǔn)識(shí)別作物邊緣特征,使邊緣線型輪廓緊貼冠層區(qū)域,減少土壤背景對(duì)棉花冠層區(qū)域提取結(jié)果的影響。Canny邊緣檢測(cè)算法剔除土壤背景效果優(yōu)于Otsu算法。圖4,圖5
2.1.2? 不同算法對(duì)冠層圖像提取精度評(píng)價(jià)
研究表明,Otsu算法的精確率大于召回率(2021年93.68%>82.52%,2022年90.42%>80.60%),Otsu算法提取冠層圖像中含有大量土壤像元,在土壤背景中包含冠層像元較少;Canny邊緣檢測(cè)算法2年準(zhǔn)確率達(dá)到了91.90%、92.76%,召回率達(dá)到了95.22%、92.98%,Canny邊緣檢測(cè)算法可較好剔除土壤背景,避免對(duì)冠層像元和土壤像元的誤分。Canny邊緣檢測(cè)算法的準(zhǔn)確率和召回率均大于Otsu算法,提取效果較優(yōu)。表1、表2
2.2? 冠層溫度提取
2.2.1? 冠層溫度直方圖
研究表明,原始熱紅外遙感圖像的溫度頻率直方圖呈雙峰變化,并經(jīng)地面實(shí)測(cè)數(shù)據(jù)驗(yàn)證,陽光直射土壤溫度高于冠層溫度,陰影土壤溫度、陰影葉片溫度低于陽光直射冠層溫度,且在太陽高度角穩(wěn)定的正午時(shí)刻陰影土壤溫度像元數(shù)量低于冠層溫度像元數(shù)據(jù)和陽光直射土壤溫度像元數(shù)量,因此,雙峰直方圖的第一個(gè)峰主要代表棉花冠層溫度信息,第二個(gè)峰主要代表陽光直射土壤溫度信息,且第二個(gè)峰的溫度大于第一個(gè)峰的溫度。棉花孕蕾期,熱紅外遙感圖像中冠層像元數(shù)量少于土壤背景像元數(shù)量,出現(xiàn)第一個(gè)峰的高度低于第二個(gè)峰,而在棉花盛蕾期,熱紅外遙感圖像中冠層像元數(shù)量多于土壤背景像元數(shù)量,使第一個(gè)峰的高度高于第二個(gè)峰。
Otsu算法、Canny邊緣檢測(cè)算法剔除土壤背景后冠層溫度頻率直方圖變化規(guī)律與原始圖像溫度頻率直方圖的第一個(gè)峰相同,均呈單峰變化。但是,Otsu算法剔除土壤背景后,冠層溫度頻率直方圖形狀粗糙,噪聲多,而Canny邊緣檢測(cè)算法剔除土壤背景后,冠層溫度頻率直方圖形狀光滑,噪聲少,且呈偏態(tài)分布。其次,Canny邊緣檢測(cè)算法提取的冠層溫度范圍小于Otsu算法提取的冠層溫度范圍。圖6,圖7
2.2.2? 冠層溫度特征值
研究表明,Otsu算法、Canny算法提取的冠層溫度最大值分別為62.40~64.69℃、59.65~63.54℃,較原始圖像冠層溫度最大值分別下降為2.98~4.11℃、5.26~5.73℃,兩算法均可降低土壤背景的干擾,但Canny算法的剔除土壤背景的效果優(yōu)于Otsu算法。原始圖像及Otsu算法、Canny算法提取的冠層溫度最小值均為11.34℃,兩算法對(duì)剔除陰影土壤溫度像元沒有影響。2種算法可以降低冠層溫度平均值,其中以Canny算法提取的冠層溫度平均值最低(29.95~30.54℃)。Canny算法較優(yōu),Otsu算法次之,原始熱紅外遙感圖像最差(4.72~11.81℃>3.64~5.85℃>2.78~3.43℃),但是Canny算法提取的冠層溫度平均值與實(shí)測(cè)溫度差值(2.78~3.43℃)仍然較大。表3、表4
2.3? 冠層溫度優(yōu)化及其與實(shí)測(cè)溫度相關(guān)性
研究表明,選擇剔除溫度頻率直方圖兩端1%溫度信息對(duì)冠層溫度進(jìn)行優(yōu)化,并將優(yōu)化前后冠層溫度與實(shí)測(cè)溫度進(jìn)行相關(guān)性分析(2021年n=30,2022年n=30)。對(duì)比Otsu算法、Canny算法,Canny算法提取的冠層溫度與實(shí)測(cè)溫度相關(guān)系數(shù)最大(0.88、0.93)、RMSE最?。?.78、2.87℃)。剔除Canny算法冠層溫度頻率直方圖兩端1%溫度信息后,冠層溫度與實(shí)測(cè)溫度相關(guān)系數(shù)最大(0.94、0.95),RMSE最?。?.59、1.43℃)。隨著土壤背景的剔除及對(duì)冠層溫度頻率直方圖的優(yōu)化,冠層溫度逐漸降低,擬合線斜率逐漸接近于1。故通過剔除Canny算法溫度頻率直方圖兩端1%溫度信息所提取的冠層溫度精度最佳,2年相關(guān)系數(shù)分別為0.94、0.95,RMSE分別為1.59、1.43℃。圖8,圖9
3? 討 論
3.1
無人機(jī)熱紅外遙感圖像中不僅包含冠層像元,還存在土壤像元,因此在提取作物冠層溫度過程中,易受土壤像元干擾,影響作物冠層溫度提取精度。為此,前人采用RGRI指數(shù)法、GBRI指數(shù)
法等剔除可見光圖像土壤背景以掩膜冠層溫度信息進(jìn)而提取冠層溫度,雖取得了較好的效果,但是,由于可見光圖像與熱紅外遙感圖像分辨率不一致,需要降低可見光圖像的分辨率,布置較多的控制點(diǎn)對(duì)熱紅外遙感圖像進(jìn)行配準(zhǔn),預(yù)處理過程復(fù)雜,不利于冠層溫度的提?。?,23-27]。研究采用Otsu算法和Canny邊緣檢測(cè)算法剔除熱紅外遙感圖像中的土壤背景,結(jié)果表明,Otsu算法剔除土壤背景后,圖像中仍含有大量的土壤與冠層混合像元,而Canny邊緣檢測(cè)算法依據(jù)作物冠層邊緣特征識(shí)別冠層區(qū)域,進(jìn)而避免更多土壤像元進(jìn)入冠層區(qū)域,2年提取棉花冠層圖像的準(zhǔn)確率達(dá)到了91.90%、92.76%,召回率達(dá)到了95.22%、92.98%,剔除土壤背景效果優(yōu)于Otsu算法,與ZHAO B[1]、張智韜等[22]在棉花、玉米上的研究結(jié)果一致。
3.2
研究表明,通過Otsu算法和Canny邊緣檢測(cè)算法剔除土壤背景后冠層溫度直方圖呈單峰偏態(tài)分布,溫度分布范圍分別為11.34~62.40℃、11.34~64.69℃(Otsu算法),11.34~59.65℃、11.34~63.54℃(Canny算法)而前人在玉米[27-28]等作物上的冠層溫度分布范圍主要在33~64℃,與研究的冠層溫度分布范圍不一致,可能是因?yàn)榍叭耸褂玫腛tsu-EXG-Kmeans算法中加入溫度閾值條件,剔除高于冠層溫度和低于冠層溫度的土壤像元,而通過Otsu算法和Canny邊緣檢測(cè)算法剔除土壤背景后,冠層溫度中會(huì)包含部分陰影土壤和陽光直射土壤,從而導(dǎo)致棉花
冠層溫度中既有低溫又有高溫,分布范圍增大。比較兩種算法下的冠層溫度頻率直方圖可以看出,使用Canny邊緣檢測(cè)算法的冠層溫度頻率直方圖形狀較Otsu算法的冠層溫度頻率直方圖形狀光滑,噪聲較少,其次Canny邊緣檢測(cè)算法提取的冠層平均溫度以及與實(shí)測(cè)溫度的差值均低于Otsu算法,Canny邊緣檢測(cè)算法提取效果較優(yōu)。研究對(duì)冠層溫度直方圖進(jìn)行優(yōu)化,并通過圖像溫度與實(shí)測(cè)溫度相關(guān)性發(fā)現(xiàn),剔除Canny邊緣檢測(cè)算法溫度直方圖兩端1%溫度信息后,圖像溫度與實(shí)測(cè)溫度相關(guān)性最大(r=0.94、0.95),擬合線接近1∶1,RMSE最小,分別為1.59、1.43℃,提取精度較高,該方法提取的冠層溫度接近實(shí)測(cè)溫度,可以較好的用于棉花水分狀況精準(zhǔn)評(píng)價(jià),但是研究選擇剔除溫度頻率直方圖兩端1%溫度信息進(jìn)行優(yōu)化,是在前人研究基礎(chǔ)上開展的,并未根據(jù)實(shí)際情況對(duì)剔除溫度直方圖兩端2%、5%的溫度信息進(jìn)行驗(yàn)證[22]。
研究中,僅針對(duì)冠層溫度提取及精度評(píng)價(jià)開展了研究,因此在確定提取熱紅外遙感圖像冠層溫度最佳方法后,下一步的研究重點(diǎn)是以無人機(jī)熱紅外遙感為基礎(chǔ),以地面?zhèn)鞲衅鳎庀?、蒸發(fā)蒸騰、土壤溫濕度等)數(shù)據(jù)為支撐,利用熱紅外無人機(jī)反演作物水分狀況數(shù)據(jù)信息對(duì)衛(wèi)星熱紅外遙感影像數(shù)據(jù)進(jìn)行校正,以實(shí)現(xiàn)在大尺度遙感反演背景下的新疆滴灌棉花水分狀況精準(zhǔn)監(jiān)測(cè)。
4? 結(jié) 論
4.1
利用Canny邊緣檢測(cè)算法剔除土壤背景效果最優(yōu),Otsu算法次之(準(zhǔn)確率為91.90%>81.52%(2021)、92.76%>80.60%)。
4.2? 與Otsu算法相比,使用Canny邊緣檢測(cè)算法提取冠層溫度頻率直方圖形狀光滑,噪聲較少,降低了冠層溫度平均值(29.95、30.54℃),且與實(shí)測(cè)溫度差值最?。?.78、3.43℃)。
4.3
剔除Canny邊緣檢測(cè)算法溫度頻率直方圖兩端1%溫度信息后,2年提取的冠層溫度與實(shí)測(cè)溫度相關(guān)性最高(r=0.94,0.95),RMSE最低(1.59、1.43℃),利用Canny邊緣檢測(cè)算法剔除土壤背景后,去除其溫度頻率直方圖兩端1%溫度信息,是準(zhǔn)確獲取冠層溫度的有效方法。
參考文獻(xiàn)(References)
[1]
ZHAO B,Adama T,Ata-UI-Karim S T,et al.Recalibrating plant water status of winter wheat based on nitrogen nutrition index using thermal images[J].Precision Agriculture, 2022,23(3):748-767.
[2]楊文攀,李長(zhǎng)春,楊浩,等.基于無人機(jī)熱紅外與數(shù)碼影像的玉米冠層溫度監(jiān)測(cè)[J].農(nóng)業(yè)工程學(xué)報(bào),2018,34(17):68-75,301.
YANG Wenpan,LI Changchun,YANG Hao,et al.Monitoring of canopy temperature of maize based on UAV thermal infrared imagery and digital imagery [J]. Transactions of the Chinese Society for Agricultural Engineering, 2018,34(17):68-75,301.
[3]蔡甲冰,許迪,司南,等.基于冠層溫度和土壤墑情的實(shí)時(shí)監(jiān)測(cè)與灌溉決策系統(tǒng)[J].農(nóng)業(yè)機(jī)械學(xué)報(bào),2015,46(12):133-139.
CAI Jiabing,XU Di,SI Nan,et al.Real-time monitoring system of crop canopy temperature and soil moisture for irrigation decision-making[J].Transactions of the Chinese Society for Agricultural Machinery,2015,46(12):133-139.
[4]史長(zhǎng)麗,郭家選,梅旭榮,等.夏玉米農(nóng)田表面溫度影響因素分析[J].中國(guó)農(nóng)業(yè)科學(xué),2006,39(1):48-56.
SHI Changli,GUO Jiaxuan,MEI Xurong,et al.Analysis of the factors influencing surface temperature in summer maize field[J].Scientia Agricultura Sinica,2006,39(1):48-56.
[5]Tanner C B.Plant temperatures [J]. Agronomy Journal, 1963,55(2):210-211.
[6]楊明欣,高鵬,陳文彬,等.基于機(jī)器學(xué)習(xí)的油青菜心水分脅迫研究[J].華南農(nóng)業(yè)大學(xué)學(xué)報(bào),2021,42(5):117-126.
YANG Mingxin,GAO Peng,CHEN Wenbin,et al.Research of Brassica chinensis var.parachinensis under water stress based on machine learning [J]. Journal of South China Agricultural University, 2021,42(5):117-126.
[7]Bian J,Zhang Z T,Chen J Y,et al.Simplified evaluation of cotton water stress using high resolution unmanned aerial vehicle thermal imagery [J].Remote Sensing, 2019,11(3):267.
[8]張宏鳴,王佳佳,韓文霆,等.基于熱紅外遙感影像的作物冠層溫度提取[J].農(nóng)業(yè)機(jī)械學(xué)報(bào),2019,50(4):203-210.
ZHANG Hongming,WANG Jiajia,HAN Wenting,et al.Crop canopy temperature extraction based on thermal infrared remote sensing images [J].Transactions of the Chinese Society for Agricultural Machinery,2019,50(4):203-210.
[9]馮子恒,宋莉,張少華,等.基于無人機(jī)多光譜和熱紅外影像信息融合的小麥白粉病監(jiān)測(cè)[J].中國(guó)農(nóng)業(yè)科學(xué),2022,55(5):890-906.
FENG Zhihong,SONG Li,ZHANG Shaohua,et al.Wheat powdery mildew monitoring based on information fusion of multi-spectral and thermal infrared images acquired with an unmanned aerial vehicle [J]. Scientia Agricultura Sinica,2022,55(5):890-906.
[10]Zhou Z,Majeed Y,Diverres Naranjo G,et al.Assessment for crop water stress with infrared thermal imagery in precision agriculture:a review and future prospects for deep learning applications [J].Computers and Electronics in Agriculture,2021,182:106019.
[11]Hou M J,Tian F,Ortega-Farias,S,et al.Estimation of crop transpiration and its scale effect based on ground and UAV thermal infrared remote sensing images [J].European Journal of Agronomy, 2021,131:126389.
[12]夏清,張振鑫,王婷婷,等.基于改進(jìn)Sobel算子的紅外圖像邊緣提取算法[J].激光與紅外,2013,43(10):1158-1161.
XIA Qing,ZHANG Zhenxin,WANG Tingting,et al.Edge extraction algorithm of infrared thermal image based on improved sobel operator [J]. Laser & Infrared, 2013,43(10):1158-1161.
[13]Liu M,Guan H O,Ma X D,et al.Recognition method of thermal infrared images of plant canopies based on the characteristic registration of heterogeneous images[J]. Computers and Electronics in Agriculture, 2020,177:105678.
[14]Zhang L Y,Niu Y X,Zhang H H,et al.Maize canopy temperature extracted from UAV thermal and RGB imagery and its application in water stress monitoring[J]. Frontiers in Plant Science,2019,10:1270.
[15]徐洪剛,陳震,程千,等.無人機(jī)熱紅外反演土壤含水率的方法[J].排灌機(jī)械工程學(xué)報(bào),2022,40(11):1181-1188.
XU Honggang,CHEN Zhen,CHENG Qian,et al.Inversion of soil moisture content based on UAV thermal infrared image [J].Journal of Drainage and Irrigation Machinery Engineering,2022,40(11):1181-1188.
[16]Sepúlveda-Reyes D,Ingram B,Bardeen M,et al.Selecting canopy zones and thresholding approaches to assess grapevine water status by using aerial and ground-based thermal imaging[J].Remote Sensing, 2016,8(10):822.
[17]張智韜,邊江,韓文霆,等.剔除土壤背景的棉花水分脅迫無人機(jī)熱紅外遙感診斷[J].農(nóng)業(yè)機(jī)械學(xué)報(bào),2018,49(10):250-260.
ZHANG Zhitao,BIAN Jiang,HAN Wenting,et al.CUI T.Diagnosis of cotton water stress using unmanned aerial vehicle thermal infrared remote sensing after removing soil background [J]. Transactions of the Chinese Society for Agricultural Machinery, 2018,49(10):250-260.
[18]張智韜,許崇豪,譚丞軒,等.覆蓋度對(duì)無人機(jī)熱紅外遙感反演玉米土壤含水率的影響[J].農(nóng)業(yè)機(jī)械學(xué)報(bào),2019,50(8):213-225.
ZHANG Zhitao,XU Chonghao,TAN Chengxuan,et al.Influence of coverage on soil moisture content of field corn inversed from thermal infrared remote sensing of UAV[J].Transactions of the Chinese Society for Agricultural Machinery,2019,50(8):213-225.
[19]程麗娜,鐘才榮,李曉燕,等.Sentinel-2密集時(shí)間序列數(shù)據(jù)和Google Earth Engine的潮間帶濕地快速自動(dòng)分類[J].遙感學(xué)報(bào),2022,26(2):348-357.
CHENG Lina,ZHONG Cairong,LI Xiaoyan,et al.Rapid and automatic classification of intertidal wetlands based on intensive time series Sentinel-2 images and Google Earth Engine[J]. National Remote Sensing Bulletin,2022,26(2):348-357.
[20]胡克滿,羅少龍,胡海燕.應(yīng)用Canny算子的織物疵點(diǎn)檢測(cè)改進(jìn)算法[J].紡織學(xué)報(bào),2019,40(1):153-158.
HU Keman,LUO Shaolong,HU Haiyan.Improved algorithm for fabric defect detection based on Canny operator[J]. Journal of Textile Research, 2019,40(1):153-158.
[21]陳玉婷,劉洞波,施怡澄,等.基于小波模極大值和自適應(yīng)Canny算子的織物疵點(diǎn)檢測(cè)算法[J].湖南工程學(xué)院學(xué)報(bào)(自然科學(xué)版),2022,32(2):54-58,65.
CHEN Yuting,LIU Dongbo,SHI Yicheng,et al.Heredity analysis and cluster structure charateristics of agcu alloy at different cooling rates and pressures[J]. Journal of Hunan Institute of Engineering(Natural Science Edition), 2022,32(2):54-58,65.
[22]張智韜,許崇豪,譚丞軒,等.基于無人機(jī)熱紅外遙感的玉米地土壤含水率診斷方法[J].農(nóng)業(yè)機(jī)械學(xué)報(bào),2020,51(3):180-190.
ZHANG Zhitao,XU Chonghao,TAN Chengxuan,et al.Diagnosing method of soil moisture content in corn field based on thermal infrared remote sensing of UAV[J]. Journal of Hnnan Institute of Engineering(Natural Science Edition),2020,51(3):180-190.
[23]Ranjan A,Sinha R,Singla-Pareek S L,et al.Shaping the root system architecture in plants for adaptation to drought stress[J].Physiologia Plantarum, 2022,174(2):e13651.
[24]Lu S,Zhang T,Tian F.Evaluation of crop water status and vegetation dynamics for alternate partial root-zone drip irrigation of alfalfa:Observation with an unmanned aerial vehicle [J]. Frontiers in Environmental Science, 843.
[25]Riveros-Burgos C,Ortega-Farías S,Morales-Salinas L,et al.Assessment of the clumped model to estimate olive orchard evapotranspiration using meteorological data and UAV-based thermal infrared imagery[J]. Irrigation Science,2021,39(1):63-80.
[26]張智韜,于廣多,吳天奎,等.基于無人機(jī)遙感影像的玉米冠層溫度提取及作物水分脅迫監(jiān)測(cè)[J].農(nóng)業(yè)工程學(xué)報(bào),2021,37(23):82-89.
ZHANG Zhitao,YU Guangduo,WU Tiankui,et al.Temperature extraction of maize canopy and crop water stress monitoring based on UAV remote sensing images[J].Transactions of the Chinese Society of Agricultural Engineering, 2021,37(23):82-89.
[27]楊帥,陳俊英,周永財(cái),等.無人機(jī)熱紅外遙感反演玉米根域土壤含水率方法研究[J].節(jié)水灌溉,2021,(3):12-18.
YANG Shuai,CHEN Junying,ZHOU Yongcai,et al.A Study on the method of UAV thermal infrared remote sensing to retrieve soil moisture content in corn root zone [J].Water Saving Irrigation,2021,(3):12-18.
Extraction and accuracy evaluation of cotton canopy temperature under drip irrigation based on uav thermal infrared remote sensing
DANG Xuwei1,LIN Xinyuan1,HE Zheng1,CHEN Yan1,CI Baoxia1,MA Xuehua1,GUO Chenli2,HE Yaxing1,LIU Yang1,3,MA Fuyu1,3
(1.The Key Laboratory of Oasis Eco-Agriculture,Xinjiang Production and Construction Corps/ College of Agriculture,Shihezi University,Shihezi Xinjiang 832003,China; 2.College of Agronomy,Nanjing Agricultural University,Nanjing 210095,China; 3.National & Local Joint Engineering Research Center of Information Management and Application Technology for Modern Agricultural Production(XPCC),Shihezi Xinjiang 832003,China)
Abstract:【Objective】 To increase the accuracy of canopy temperature extraction derived from thermal infrared imagery of drip irrigated cotton in Xinjiangin the hope of providing a technical support for precise water status monitoring.
【Methods】 Different soil moisture contents were set at cotton seedling and squaring stages.The thermal infrared images of different treatments were acquired by using UAV,and the temperature of radiation calibration plate in plot and water in the bucket were measured by using a portable handheld thermometer.For the above information,the latter temperature was used to calibrate the former temperature extracted from thermal imagery.The Otsu and Canny edge detection algorithms were used to mask thermal infrared images and remove soil background.Cotton canopy temperature was extracted by region of interest(ROI) and band math,and then the canopy temperature frequency histograms were plotted and optimized.Meanwhile,the actual cotton canopy temperature was obtained from a portable handheld thermometer.The consistency analysis was performed between actual canopy temperature and extracted canopy temperature to calibrate the accuracy of extracted temperature from thermal imagery.
【Results】 Canny edge detection algorithm eliminated soil background and extracted canopy image with higher accuracy than Otsu algorithm(91.90%>82.52%、92.76%>80.60%),which reached the best effect.The canopy temperature histograms constructed by Otsu algorithm and Canny edge detection algorithm after removing soil background are skewed,but the canopy temperature histograms constructed by Canny edge detection algorithm after removing soil background were smoother and less noisy than Otsu algorithm,and the average canopy temperature of Canny edge detection algorithm in two years was the lowest(29.95,30.54℃),with the smallest difference from the measured temperature(2.78,3.43℃).Correlation analysis showed that the extracted canopy temperature based on Canny edge detection algorithm had the highest correlation with the measured temperature(r=0.94,0.95) and the lowest RMSE(1.59,1.43℃),where the 1% temperature information at both ends of the temperature histogram of Canny edge detection algorithm was dislogded.
【Conclusion】? The Canny edge detection algorithm improves the precision of cotton canopy temperature extraction from UAV thermal infrared images,and the optimization of 1% temperature at both ends of the temperature histogram is helpful to improve the precision of cotton canopy temperature extraction.
Key words:drip-irrigated cotton; unmanned aerial vehicle; thermal infrared image; canopy temperature; accuracy
Fund projects:Xinjiang Production and Construction Corps Financial Science and Technotogy Plan Project“Key Cultivation Technology Integration and Demonstration for Xinjiang High-Quality Cotton Production”(2020AB017); “Development and Application Demonstration of Main Crop Precision Water and Fertilizer Integration Technology and Equipment Research”(2020AB018)
Correspondence author:LIU Yang(1989-),male,from Qinghai,associate professor,doctoral student,research field:efficient use of water and fertilizer and precision cultivation of crops,(E-mail)ly.0318@163.com
MA Fuyu(1967-),male,from Gansu,professor,doctoral student,research field:crop precision cultivation and water and fertilizer integration,(E-mail)1469633844@qq.com