賈桂鋒,蒙俊宇,武 墩,王登輝,高 云,馮耀澤
?
被毛對(duì)熱成像檢測(cè)生豬體表溫度精度的影響及噪聲濾除方法
賈桂鋒1,2,蒙俊宇1,武 墩1,王登輝1,高 云1,2,馮耀澤1,2※
(1. 華中農(nóng)業(yè)大學(xué)工學(xué)院,武漢 430070;2. 農(nóng)業(yè)部長(zhǎng)江中下游農(nóng)業(yè)裝備重點(diǎn)實(shí)驗(yàn)室,武漢 430070)
生豬皮膚的溫度分布是表征其生理狀態(tài)和疾病的重要指標(biāo),通常由紅外熱成像技術(shù)(infrared thermography, IRT)檢測(cè),然而由于生豬體表附有被毛在熱圖像中產(chǎn)生大量的溫度噪聲,降低了IRT對(duì)皮膚溫度的檢測(cè)精度。該文針對(duì)此問(wèn)題探索被毛對(duì)皮膚溫度分布的影響規(guī)律,并設(shè)計(jì)消除被毛影響的熱圖像降噪算法,提高對(duì)溫度分布的檢測(cè)精度。通過(guò)對(duì)12頭生豬試驗(yàn),分析目標(biāo)區(qū)域在正常被毛和剔除被毛后溫度分布的統(tǒng)計(jì)量得出被毛在溫度分布中產(chǎn)生大量的“峽谷”狀低溫噪聲,顯著降低了目標(biāo)區(qū)域的最低溫度及平均溫度。根據(jù)毛發(fā)噪聲的影響規(guī)律提出網(wǎng)格化最大值-雙三次插值算法并確定算法的最佳鄰域尺寸為4.25mm。采用均方誤差、峰值信噪比等指標(biāo)定量評(píng)價(jià)算法的有效性,結(jié)果表明經(jīng)算法處理后,均方誤差由0.38下降到0.05(<0.01),峰值信噪比由45.14 dB上升到53.66 dB(<0.01),說(shuō)明該算法能夠?yàn)V除熱圖像中毛發(fā)引起的噪聲,可提高IRT對(duì)溫度分布的檢測(cè)精度。
紅外熱成像;溫度分布;濾波器;豬;圖像插值;算法
生豬體表的溫度分布是表征其生理狀態(tài)和疾病的重要指標(biāo),可用于異常行為識(shí)別[1]、發(fā)育狀況評(píng)估[2]、炎癥檢測(cè)[3-4]、排卵預(yù)測(cè)[5-6]及發(fā)熱診斷[7-8]。特別在發(fā)熱診斷應(yīng)用中,通過(guò)溫度異常可早期檢測(cè)豬瘟、偽狂犬病、藍(lán)耳、圓環(huán)和豬肺疫等伴隨有發(fā)熱癥狀的主要流行病[9],避免造成呼吸、消化和繁殖障礙[10]。紅外熱成像(Infrared thermography, IRT)測(cè)溫技術(shù)因其非接觸式、靈敏度高且響應(yīng)時(shí)間快等優(yōu)點(diǎn)在畜牧動(dòng)物檢測(cè)中備受關(guān)注,且不會(huì)造成應(yīng)激反應(yīng)等負(fù)面影響[11-13]。IRT測(cè)溫原理是利用非制冷紅外探測(cè)器捕獲動(dòng)物體表輻射的長(zhǎng)波段紅外線(波長(zhǎng)范圍通常為7~13m)并將輻射強(qiáng)度按一定的空間分辨率轉(zhuǎn)換為數(shù)字圖像[14],以反映體表各點(diǎn)溫度的高低與分布,繼而揭示動(dòng)物的生理狀態(tài)及健康狀況[3,15]。Sapkota等在研究豬的體表溫度與核心溫度之間的關(guān)系時(shí)將肩部、胸部、臀部等9個(gè)區(qū)域的平均溫度作為體表溫度以研究其能否準(zhǔn)確反映體溫的變化[16];Menzel等根據(jù)胸部生理結(jié)構(gòu)在第5、7、10根胸椎處各選取3個(gè)直徑1 cm的圓形區(qū)域和腹部2個(gè)較大區(qū)域作為目標(biāo)區(qū)域(regions of interest, ROI),再根據(jù)這些ROI熱成像的最高溫度和平均溫度與肺部的計(jì)算機(jī)斷層掃描所測(cè)量的組織厚度進(jìn)行相關(guān)性分析,研究胸部溫度分布與胸部及肺部組織厚度的關(guān)系[17]。通常生豬體表有不同程度的被毛附著,遮擋皮膚的熱輻射,影響體表溫度的準(zhǔn)確提取,甚至使之不能用于發(fā)熱診斷[18]。本文針對(duì)該問(wèn)題基于紅外熱成像技術(shù)探索生豬被毛對(duì)體表溫度檢測(cè)精度的影響,并根據(jù)影響規(guī)律提出被毛噪聲的濾除算法,以提高表面溫度分布檢測(cè)精度和疾病診斷能力。
試驗(yàn)數(shù)據(jù)于2018年8月在安徽省臨泉縣某生豬養(yǎng)殖場(chǎng)獲取,來(lái)源于12頭處于空懷期的母豬,胎次2~3胎。環(huán)境溫濕度用環(huán)境指標(biāo)測(cè)量?jī)x(Victor,VC231)測(cè)量,試驗(yàn)中豬舍內(nèi)的平均環(huán)境溫度和濕度分別為27.4 ℃和80.3%。體表溫度采用手持式紅外熱像儀(Fluke,Ti-300)檢測(cè),該儀器分辨率為240×180像素,靈敏度達(dá)50mK,精度為2%,可同時(shí)采集相同區(qū)域的可見(jiàn)光圖像,并搭載精度0.01 m的激光測(cè)距傳感器,用于測(cè)量被測(cè)目標(biāo)到熱像儀的距離。
試驗(yàn)方案是通過(guò)對(duì)比感興趣區(qū)域(region of interest, ROI)內(nèi)正常被毛(normal coat,NC)和剔除被毛后(shed coat,SC)兩種狀態(tài)下熱成像的測(cè)溫?cái)?shù)據(jù),分析被毛對(duì)IRT測(cè)溫的影響,根據(jù)影響規(guī)律設(shè)計(jì)被毛噪聲的濾除算法并驗(yàn)證算法有效性。在生豬背部最后一根肋骨距中心線約6 cm處選取5 cm×5 cm的方形區(qū)域作為ROI并用記號(hào)筆標(biāo)記,該區(qū)域較為平坦且易于辨識(shí),可確保各試驗(yàn)動(dòng)物選取區(qū)域的一致性,故選擇該位置作為ROI。IRT測(cè)量前將熱像儀的發(fā)射率設(shè)定為0.97[19-20],背景溫度設(shè)置為當(dāng)前環(huán)境溫度,溫度測(cè)量范圍設(shè)定為?20~80 ℃。就緒后采用熱像儀在ROI上方0.3 m處采集NC狀態(tài)下的熱圖像,然后立即用刀具將ROI內(nèi)的毛發(fā)剃除干凈,在同樣方位和距離下采集SC狀態(tài)下的熱圖像,共采集24幅熱圖像,用于提取ROI在兩種狀態(tài)下的溫度分布,以分析被毛對(duì)測(cè)溫精度的影響。
熱圖像的溫度數(shù)據(jù)由SmartView(Fluke Co, Ltd)軟件提取,用矩形選取工具選擇熱圖像的ROI區(qū)域,分別提取區(qū)域中的最大值、最小值、平均值和溫度矩陣的標(biāo)準(zhǔn)差,共12組數(shù)據(jù),統(tǒng)計(jì)結(jié)果見(jiàn)表1。
表1 NC和SC狀態(tài)下IRT測(cè)溫的統(tǒng)計(jì)數(shù)據(jù) Table 1 Statistical temperature of ROI(region of interest) measured by IRT in NC (normal coat) and SC (shed coat) status ℃
注:NC表示ROI正常被毛狀態(tài)下的熱圖像,SC表示ROI被毛剔除狀態(tài)下的熱圖像。
Note:NCindicates the thermal image of ROI under the normal coat state , andSCindicates the thermal image of ROI under shed coat status.
從數(shù)據(jù)統(tǒng)計(jì)結(jié)果中可看出:1)ROI區(qū)域內(nèi)毛發(fā)剃除前后,IRT所測(cè)溫度的統(tǒng)計(jì)值均存在差異,SC狀態(tài)下的溫度最大值、最小值和平均值分別比NC狀態(tài)下平均增大0.19、1.59和0.47 ℃,而標(biāo)準(zhǔn)差減小0.22 ℃,且檢驗(yàn)表明2種狀態(tài)下溫度的最小值、標(biāo)準(zhǔn)差存在非常顯著的差異(<0.01),而最大值和平均值差異不顯著(>0.05),說(shuō)明被毛對(duì)紅外測(cè)溫的最小值和標(biāo)準(zhǔn)差存在顯著影響,對(duì)平均值有一定的影響,而對(duì)最大值則影響較弱;2)標(biāo)準(zhǔn)差反映了數(shù)據(jù)的離散程度,正常被毛時(shí)溫度數(shù)據(jù)的離散程度較大,而剔除被毛后數(shù)據(jù)的離散程度顯著減小,同時(shí)最小值和平均值增大,說(shuō)明被毛引入了大量的低溫噪聲。通過(guò)觀察NC狀態(tài)下的熱圖像(圖1)也發(fā)現(xiàn)最小值均是由毛發(fā)引起的,并非是真實(shí)的皮膚溫度;3)正常被毛時(shí)的最高溫度比無(wú)被毛時(shí)略低0.19 ℃,在疾病診斷的允許誤差范圍±0.3 ℃(經(jīng)驗(yàn)數(shù)據(jù))內(nèi),仍具有診斷意義,故認(rèn)為最高溫度客觀地反映了ROI內(nèi)的皮膚溫度。
因此,正常被毛時(shí)IRT檢測(cè)的溫度分布存在大量低溫噪聲,不能直接用于生理狀況評(píng)估與診斷,而如何將體表熱圖像通過(guò)圖像處理濾除毛發(fā)噪聲以客觀地表征皮膚的溫度分布是準(zhǔn)確評(píng)估生豬發(fā)熱狀態(tài)、疾病診斷等應(yīng)用的關(guān)鍵問(wèn)題。
生豬背部ROI區(qū)域在毛發(fā)剃除前后的溫度分布分別如圖1a和1b所示。由圖1a可知,NC狀態(tài)下的溫度分布面上隨機(jī)出現(xiàn)了若干條“峽谷”狀的凹陷,這些相互交錯(cuò)的凹陷降低了ROI的平均溫度和最低溫度,影響皮膚溫度的檢測(cè)。毛發(fā)引起的噪聲不同于高斯、瑞利等服從特定分布的概率密度噪聲[21],難以用數(shù)學(xué)模型描述和處理,該噪聲由被測(cè)物的結(jié)構(gòu)引起,可稱(chēng)之為結(jié)構(gòu)型噪聲,且噪聲紋理具有一定的方向性。研究該類(lèi)型噪聲的濾波方法需根據(jù)其分布特點(diǎn)展開(kāi)。圖1b所示的是SC狀態(tài)下溫度分布,其溫度變化較為平緩,未出現(xiàn)“峽谷”狀的溫度凹陷等噪聲,客觀地表征了皮膚表面的溫度分布,可作為評(píng)價(jià)濾波算法的真實(shí)溫度分布。
圖1 NC和SC狀態(tài)下IRT檢測(cè)的ROI溫度分布
圖像濾波通常是選擇圖像中的一點(diǎn)并將該點(diǎn)鄰域×像素內(nèi)的數(shù)據(jù)點(diǎn)與濾波模板進(jìn)行運(yùn)算,運(yùn)算結(jié)果為該點(diǎn)的響應(yīng),對(duì)圖像采用滑動(dòng)鄰域操作或分離鄰域操作即可對(duì)整幅圖像處理[22-23]。為準(zhǔn)確提取生豬熱圖像中的皮膚溫度值,應(yīng)避免由毛發(fā)引起的低溫噪聲參與運(yùn)算,本文根據(jù)被毛對(duì)IRT測(cè)溫值的影響規(guī)律及噪聲結(jié)構(gòu)提出一種網(wǎng)格化最大值-雙三次插值算法(grid maximum- bicubic interpolation, GMBI),其流程如下:
1)圖像網(wǎng)格化分割,首先設(shè)置鄰域尺寸,然后將圖像分割為若干個(gè)方塊,即、的大小均為,分割時(shí)要求每個(gè)塊內(nèi)至少包含1個(gè)皮膚溫度數(shù)據(jù)。若鄰域尺寸過(guò)小則圖像塊內(nèi)可能完全被噪聲覆蓋,無(wú)法提取皮膚溫度,而尺寸過(guò)大時(shí)則降低熱圖像對(duì)皮溫分布的分辨力,故鄰域尺寸的確定較為關(guān)鍵。
2)根據(jù)毛發(fā)對(duì)體表溫度的影響規(guī)律可知體表的最高溫度能客觀反映皮膚溫度,故搜索各圖像塊內(nèi)的最高溫度作為相應(yīng)網(wǎng)格的響應(yīng)值,得到圖像。
3)對(duì)圖像通過(guò)二維插值算法重建熱圖像,常用的內(nèi)插核有盒狀核、三角核和立方核,其中立方核生成的曲面具有連續(xù)的二階導(dǎo)數(shù)和最小的平方曲率,灰度變化較為平滑[24],與溫度分布特征相吻合,故采用立方核進(jìn)行圖像重建。設(shè)重建圖像的點(diǎn)()在中的映射為(00),則采用該點(diǎn)4×4鄰域內(nèi)16個(gè)點(diǎn)(x,y)的值作三次插值,再由式(1)計(jì)算16個(gè)像素點(diǎn)的權(quán)重。
式中為(x,y)到(00)的距離,()為該距離下對(duì)應(yīng)的權(quán)重,常量影響清晰度,通常取-0.5[25]。由于圖像是二維矩陣,橫向和縱向分別進(jìn)行插值運(yùn)算。然后再與(x,y)按式(2)做卷積運(yùn)算。
得到的(,)即為重建圖像點(diǎn)(,)的值,迭代計(jì)算,即可重建整個(gè)圖像[26]。
4)輸出重建后的圖像。
GMBI濾波算法實(shí)現(xiàn)的關(guān)鍵問(wèn)題是如何選擇合適的鄰域尺寸以滿(mǎn)足網(wǎng)格化后每個(gè)塊內(nèi)至少包含1個(gè)皮膚溫度值,且分辨率盡可能高。實(shí)際上,由于被毛厚度在不同部位上分布不均且豬只之間也存在差異,因此難以直接確定鄰域尺寸。本文采用數(shù)理統(tǒng)計(jì)的方法在同一養(yǎng)殖區(qū)內(nèi)隨機(jī)選取另外的18頭母豬在NC狀態(tài)下采集相同ROI的熱圖像進(jìn)行研究,具體計(jì)算方法采用不同的鄰域尺寸對(duì)熱圖像進(jìn)行處理,圖像分辨率為×,鄰域尺寸L在2到/4范圍內(nèi)以1為步長(zhǎng)增長(zhǎng),GMBI濾波算法處理后得到重建圖像I,并以I、I+1之間的均方誤差(mean square error,MSE)為準(zhǔn)則進(jìn)行判別,MSE()計(jì)算方法見(jiàn)式(3)。
式中為圖像分辨率,為循環(huán)序數(shù),I指鄰域尺寸為L時(shí)算法處理得到的重建圖像。故每幅熱圖像經(jīng)迭代處理后會(huì)得到一條MSE()關(guān)于L的關(guān)系曲線,為刻畫(huà)MSE()與L的內(nèi)在關(guān)系,對(duì)18組MSE()數(shù)據(jù)進(jìn)行曲線擬合,并用FCmse表示,以刻畫(huà)MSE()隨L的變化規(guī)律。定義FCmse的累計(jì)貢獻(xiàn)率(accumulative contribution rate,ACR)用于表征不同鄰域尺寸L對(duì)圖像的消噪能力[27],ACR越大,其對(duì)應(yīng)鄰域尺寸L的消噪能力越強(qiáng),ACR計(jì)算方法如式(4),各指標(biāo)其隨鄰域尺寸L的變化曲線如圖2所示。
注:鄰域尺寸為L(zhǎng)×L像素。
由圖2可知:1)在鄰域尺寸較小時(shí),I、I-1之間的MSE較大,隨著鄰域尺寸的增大,I、I-1之間的MSE急劇減小,而后趨于平緩。由此說(shuō)明,隨著鄰域尺寸的增大,算法對(duì)熱圖像濾波品質(zhì)的改善能力逐漸減?。?)從貢獻(xiàn)率曲線可得出當(dāng)鄰域尺寸為8個(gè)像素點(diǎn)時(shí)對(duì)應(yīng)的貢獻(xiàn)率為0.92,說(shuō)明此時(shí)已消除92%的噪聲,而大于此鄰域尺寸時(shí),噪聲的濾除能力難以進(jìn)一步改善,反而降低了對(duì)皮表溫度分布的分辨力。由于像素對(duì)應(yīng)的實(shí)際尺寸由熱像儀的視場(chǎng)角和拍攝距離決定,試驗(yàn)中的拍攝距離為300 mm,根據(jù)視場(chǎng)的幾何關(guān)系可算出8個(gè)像素對(duì)應(yīng)的實(shí)際尺寸是4.25 mm。綜上所述,4.25 mm為熱圖像處理的最佳鄰域尺寸,此時(shí)既濾除低溫噪聲,還對(duì)溫度分布具有較高的分辨力。圖3展示了鄰域尺寸為2.13 mm(=4像素),4.25 mm(=8像素)和10.63 mm(=20像素)時(shí)GMBI算法對(duì)圖1a熱圖像的處理結(jié)果。
由圖3可知,當(dāng)=4像素時(shí),重建圖像的溫度分布尚存在一定的噪聲,起伏紊亂;當(dāng)=8像素時(shí),溫度分布的變化趨于平緩,與圖1b中無(wú)毛發(fā)時(shí)的溫度分布相吻合;當(dāng)=20像素時(shí),鄰域尺寸過(guò)大,雖噪聲被濾除,但溫度變化較平,忽略較多細(xì)節(jié),降低了溫度分布的分辨力。由此表明4.25 mm是GMBI算法消除毛發(fā)噪聲的最佳鄰域尺寸。
圖3 不同鄰域尺寸對(duì)熱圖像噪聲濾除效果的對(duì)比
GMBI算法的有效性采用文中12頭生豬的ROI熱圖像數(shù)據(jù)進(jìn)行驗(yàn)證。為定量評(píng)價(jià)算法的有效性,分別采用NC狀態(tài)下的熱圖像和GMBI算法重建的圖像與SC熱圖像間的MSE、峰值信噪比(peak signal-to-noise ratio,PSNR)及統(tǒng)計(jì)量差值作為算法性能的評(píng)價(jià)指標(biāo)[24,28]。MSE用于度量?jī)煞鶊D像之間的差異程度,PSNR表示信號(hào)的最大功率與噪聲功率的比值,常用于評(píng)價(jià)圖像的重建質(zhì)量[29],MSE和PSNR的計(jì)算方法分別見(jiàn)式(5)和式(6)[30]。統(tǒng)計(jì)量差值則是指兩圖像溫度數(shù)據(jù)的最大值、最小值和平均值的差值。
式中為降噪后的重建熱像,分辨率為×像素,SC為SC狀態(tài)下的熱圖像,MAX表示熱圖像測(cè)溫的量程,試驗(yàn)中設(shè)置的測(cè)溫范圍為-20~80℃,故MAX的值為100。對(duì)熱圖像數(shù)據(jù)計(jì)算后的各個(gè)指標(biāo)見(jiàn)表2,其中N-S對(duì)應(yīng)的數(shù)據(jù)是指NC熱圖像與SC熱圖像之間的參數(shù)指標(biāo),G-S指經(jīng)GMBI重建的圖像與SC熱圖像之間的參數(shù)指標(biāo)。
表2 GMBI算法對(duì)NC狀態(tài)熱圖像處理前后的試驗(yàn)結(jié)果
注:N-S指NC狀態(tài)下的熱圖像與SC狀態(tài)下的熱圖像之間的參數(shù)指標(biāo),G-S指NCNC狀態(tài)下的熱圖像經(jīng)GMBI算法重建后與SC狀態(tài)下的熱圖像之間的參數(shù)指標(biāo)。
Note:N-Srefers to the parameter between the NC thermal image and the SC thermal image,G-Smeans the parameter between the image reconstructed by the GMBI and the SC thermal image.
根據(jù)表2數(shù)據(jù)可得出:1)算法處理后的熱圖像與SC熱圖像的最小值差和平均值差比處理前顯著減?。?0.01),分別由原來(lái)的1.59 ℃、0.47 ℃下降到0.13 ℃和0.07 ℃,最大值變化不大,同時(shí)MSE由0.38顯著下降到0.05(<0.01)。這些變化說(shuō)明GMBI算法處理后熱圖像的統(tǒng)計(jì)指標(biāo)已接近無(wú)毛發(fā)狀態(tài)下的熱圖像,能客觀表征生豬表面的溫度分布。2)PSNR指標(biāo)的均值由45.14 dB顯著上升到53.66 dB(<0.01),說(shuō)明經(jīng)算法處理后熱圖像的信噪比得到非常顯著的提升,即已消除毛發(fā)引起的低溫噪聲,改善了重建圖像的質(zhì)量。
本文通過(guò)熱成像數(shù)據(jù)分析確定了生豬體表毛發(fā)對(duì)溫度檢測(cè)精度的影響規(guī)律,并根據(jù)影響規(guī)律設(shè)計(jì)熱圖像處理算法以消除毛發(fā)引起的噪聲。
1)通過(guò)熱成像數(shù)據(jù)分析,證實(shí)了基于紅外熱成像測(cè)溫時(shí)毛發(fā)對(duì)生豬皮膚表面溫度的分布存在影響,毛發(fā)產(chǎn)生大量的低溫噪聲顯著地降低了ROI的最低溫度(<0.01),而對(duì)最高溫度影響較小(>0.05),最高溫度仍能客觀地反映皮膚溫度,具有診斷意義。
2)根據(jù)毛發(fā)對(duì)熱圖像的影響規(guī)律提出網(wǎng)格化最大值-雙三次插值算法,采用不同鄰域尺寸熱圖像進(jìn)行迭代處理,并以均方誤差累計(jì)貢獻(xiàn)率為準(zhǔn)則確定算法的最佳鄰域尺寸為4.25 mm。
3)采用經(jīng)網(wǎng)格化最大值-雙三次插值算法(grid maximum -bicubic interpolation, GMBI)處理前后的正常被毛熱圖像與無(wú)被毛熱圖像之間的均方誤差MSE、峰值信噪比PSNR及統(tǒng)計(jì)量差值定量評(píng)價(jià)算法的有效性,結(jié)果表明MSE由處理前的0.38顯著下降到處理后0.05(<0.01),PSNR由45.14 dB顯著上升到53.66 dB(<0.01),說(shuō)明提出的GMBI算法是正確的,可顯著提高圖像信噪比,改善重建圖像的質(zhì)量,使得NC熱圖像準(zhǔn)確表征皮膚的溫度分布。
[1] Cook N J, Bench C J, Liu T, et al. The automated analysis of clustering behaviour of piglets from thermal images in response to immune challenge by vaccination[J]. Animal, 2018, 12(1): 122-133.
[2] Caldara F R, dos Santos L S, Machado S T, et al. Piglets’ surface temperature change at different weights at Birth[J]. Asian-Australasian Journal of Animal Sciences, 2014, 27(3): 431-438.
[3] Sathiyabarathi M, Jeyakumar S, Manimaran A, et al. Infrared thermography: A potential noninvasive tool to monitor udder health status in dairy cows[J]. Veterinary World, 2016, 9(10): 1075-1081.
[4] Menzel A, Beyerbach M, Siewert C, et al. Actinobacillus pleuropneumoniae challenge in swine: Diagnostic of lung alterations by infrared thermography[J]. BMC Veterinary Research, 2014, 10(1): 199.
[5] Simoes V G, Lyazrhi F, Picard-Hagen N, et al. Variations in the vulvar temperature of sows during proestrus and estrus as determined by infrared thermography and its relation to ovulation[J]. Theriogenology, 2014, 82(8): 1080-1085.
[6] Luno V, Gil L, Jerez R A, et al. Determination of ovulation time in sows based on skin temperature and genital electrical resistance changes[J]. Veterinary Record, 2013, 172(22): 579.
[7] Traulsen I, Naunin K, Mueller K, et al. Application of infrared thermography to measure body temperature of sows[J]. Zuchtungskunde, 2010, 82(6): 437-446.
[8] Chung T H, Jung W S, Nam E H, et al. Comparison of rectal and infrared thermometry for obtaining body temperature of gnotobiotic piglets in conventional portable germ free facility[J]. Asian-Australasian Journal of Animal Sciences, 2010, 23(10): 1364-1368.
[9] 李春華,王英,蔣鳳英,等.豬偽狂犬病研究進(jìn)展[J]. 動(dòng)物醫(yī)學(xué)進(jìn)展,2008,29(3):68-72.
Li Chunhua, Wang Ying, Jiang Fengying, et al. Progress on porcine pseudorabies[J]. Progress in Veterinary Medicine, 2008, 29(3): 68-72. (in Chinese with English abstract)
[10] 陳煥春. 豬系統(tǒng)性疾病的流行現(xiàn)狀與防控措施[J]. 飼料與畜牧,2018(2):45-50.
[11] Stewart M, Webster J R, Schaefer A L, et al. Infrared thermography as a non-invasive toot to study animal welfare[J]. Animal Welfare, 2005, 14(4): 319-325.
[12] Petry A, McGilvray W, Rakhshandeh A R, et al. Technical note: Assessment of an alternative technique for measuring body temperature in pigs[J]. Journal of Animal Science, 2017, 95(7): 3270-3274.
[13] 曹哲,施正香,安欣,等. 基于熱成像技術(shù)的牛舍圍護(hù)結(jié)構(gòu)傳熱阻測(cè)試方法[J]. 農(nóng)業(yè)工程學(xué)報(bào),2017,33(24):235-241.
Cao Zhe, Shi Zhengxiang, An Xin, et al. Evaluation on measure method of heat transfer resistance for enveloped structure of cattle barn based on infrared imaging method[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(24): 235-241. (in Chinese with English abstract)
[14] Usamentiaga R, Venegas P, Guerediaga J, et al. Infrared thermography for temperature measurement and non-destructive testing[J]. Sensors, 2014, 14(7): 12305-12348.
[15] Lahiri B B, Bagavathiappan S, Jayakumar T, et al. Medical applications of infrared thermography: A review[J]. Infrared Physics & Technology, 2012, 55(4): 221-235.
[16] Sapkota A, Herr A, Johnson J S, et al. Core body temperature does not cool down with skin surface temperature during recovery at room temperature after acute heat stress exposure[J]. Livestock Science, 2016, 191: 143-147.
[17] Menzel A, Siewert C, Gasse H, et al. Infrared thermography of the pig thorax: An assessment of selected regions of interest by computed tomographical and anatomical parameters[J]. AnatomiaHistologiaEmbryologia, 2015, 44(2): 107-117.
[18] Bekkering J, Hoy S. Continuous monitoring of ear temperature in boars[J]. Dtw Deutsche Tierarztliche Wochenschrift, 2007, 114(1): 16-19.
[19] Zhang K, Jiao L, Zhao X, et al. An instantaneous approach for determining the infrared emissivity of swine surface and the influencing factors[J]. Journal of Thermal Biology, 2016, 57: 78-83.
[20] Soerensen D D, Clausen S, Mercer J B, et al. Determining the emissivity of pig skin for accurate infrared thermography[J]. Computers and Electronics in Agriculture, 2014, 109: 52-58.
[21] 賈偉寬,趙德安,阮承治,等. 蘋(píng)果采摘機(jī)器人夜間圖像降噪算法[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(10):219-226.
Jia Weikuan, Zhao Dean, Ruan Chengzhi, et al. De-noising algorithm of night vision image for apple harvesting robot[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(10): 219-226. (in Chinese with English abstract)
[22] 王海超,王春光,宗哲英,等. 基于噪聲類(lèi)型及強(qiáng)度估計(jì)的狹葉錦雞兒葉切片圖像盲去噪[J]. 農(nóng)業(yè)工程學(xué)報(bào),2017,33(10):229-238.
Wang Haichao, Wang Chunguang, Zong Zheying, et al. Blind image denoising of microscopic slices image of Caraganastenophylla Pojark based on noise type and intensity estimation[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2017, 33(10): 229-238. (in Chinese with English abstract)
[23] 廖建尚,王立國(guó),郝思媛. 基于雙邊濾波和空間鄰域信息的高光譜圖像分類(lèi)方法[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2017,48(8):140-146,211.
Liao Jianshang, Wang Liguo, Hao Siyuan. Hyperspectral image classification method combined with bilateral filtering and pixel neighborhood information[J]. Transactions of the Chinese Society for Agricultural Machinery, 2017, 48(8): 140-146, 211.(in Chinese with English abstract)
[24] 鐘寶江,陸志芳,季家歡. 圖像插值技術(shù)綜述[J]. 數(shù)據(jù)采集與處理,2016,31(6):1083-1096.
Zhong Baojiang, Lu Zhifang, Ji Jiahuan. Review on image interpolation techniques[J]. Journal of Data Acquisition and Processing, 2016, 31(6): 1083-1096. (in Chinese with English abstract)
[25] Keys R G. Cubic convolution interpolation for digital image processing[J]. IEEE Transactions on Acoustics Speech and Signal Processing, 1981, 29(6): 1153-1160.
[26] 張玉存,張喜英,付獻(xiàn)斌,等. 基于小波與雙三次插值的高溫鍛件紅外圖像增強(qiáng)方法[J]. 中國(guó)機(jī)械工程,2017,28(17):2095—2099.
Zhang Yucun, Zhang Xiying, Fu Xianbin, et al. Infrared image enhancement algorithm for hot forgings based on wavelet transform and bicubic interpolation[J]. China Mechanical Engineering, 2017, 28(17): 2095-2099. (in Chinese with English abstract)
[27] 王麗杰,楊羽翼,代敏,等. 基于直方圖分層映射的近紅外光譜預(yù)處理算法[J]. 激光與光電子學(xué)進(jìn)展,2017,54(9):393-401.
Wang Lijie, Yang Yuyi, Dai Min, et al. Near infrared spectral pre-processing algorithm based on histogram layering mapping[J]. Laser & Optoelectronics Progress, 2017, 54(9): 393-401. (in Chinese with English abstract)
[28] 劉姍姍,白美健,許迪,等. 畦田灌溉模擬中田面微地形空間分布插值方法改進(jìn)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(17):108-114.
Liu Shanshan, Bai Meijian, Xu Di, et al. Improvement of interpolation methods for surface micro-topography spatial distribution in border irrigation simulation[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(17): 108-114. (in Chinese with English abstract)
[29] 黃小喬,石俊生,楊健,等. 基于色差的均方誤差與峰值信噪比評(píng)價(jià)彩色圖像質(zhì)量研究[J]. 光子學(xué)報(bào),2007,36(S1):295-298.
Huang Xiaoqiao, Shi Junsheng, Yang Jian, et al. Study on color image quality evaluation by MSE and PSNR based on color difference[J]. Acta Photonica Sinica, 2007, 36(S1): 295-298. (in Chinese with English abstract)
[30] 肖祥元,景文博,趙海麗. 基于峰值信噪比改進(jìn)的圖像增強(qiáng)算法[J]. 長(zhǎng)春理工大學(xué)學(xué)報(bào):自然科學(xué)版,2017,40(4):83-86,92.
Xiao Xiangyuan, Jing Wenbo, Zhao Haili. An improved image enhancement algorithm based on the peak-signal to noise ratio[J]. Journal of Changchun University of Science and Technology: Natural Science Edition, 2017, 40(4): 83-86, 92. (in Chinese with English abstract)
Effect of hair on thermometry of skin by infrared thermography and noise reduction method for live pigs
Jia Guifeng1,2, Meng Junyu1, Wu Dun1, Wang Denghui1, Gao Yun1,2, Feng Yaoze1,2※
(1.,,430070,; 2.,,430070,)
The temperature distribution of pig skin is an important indicator to characterize its physiological state and disease. However, due to the surface hair coat, the skin temperature accuracy which detected by infrared thermography (IRT) is affected and its ability to diagnosis of fever and disease is reduced. The purpose of this paper is to explore the influence patterns of the coat on the skin temperature distribution and propose the thermal image processing method to eliminate the influence of the coat on temperature accuracy. The animals for experimental data were 12 sows in empty pregnant period with the average ambient temperature of 27.4 ℃ and humidity in the piggery of 80.3% respectively. The body surface temperature was measured by hand-held infrared thermal imager (Fluke, Ti 300) with a resolution of 240 pixels×180 pixels and sensitivity of 50 mK. And it also carried a laser distance measuring sensor with a resolution of 0.01 m to measure the distance between the measured object and the thermal imager. The statistics of the temperature distribution detected by IRT from the region of interest (ROI) under normal coat (NC) was compared to that under shed coat (SC) state. The statistical data indicated that the hair coat produced a large number of “canyon”-like low temperature noise in temperature distribution in NC state, which reduced the minimum temperature and average temperature of the ROI, but had no significant effect on the maximum temperature with diagnostic ability. According to the noise distribution characteristics and the influence pattern, an image noise filtering algorithm named the grid maximum value bicubic interpolation filter (GMBI) was proposed. The GMBI algorithm consisted of three steps including image mesh segmentation, filtering with maximum value and image bicubic interpolation. The key problem of GMBI was how to select the appropriate neighborhood size to ensure that each block contained at least one skin temperature value and the resolution was as high as possible. In this study, mathematical statistics was employed and it was found out that the optimal neighborhood size was 4.25 mm. In order to evaluate the validity of the algorithm quantitatively, the mean square error (MSE), peak signal-to-noise ratio (PSNR) and the difference of maximum, minimum and mean between the processed images by GMBI and the SC thermal images were calculated. The experimental data showed that the differences of minimum and average were greatly reduced from the original 1.59 and 0.47 to 0.13 and 0.07 ℃ (<0.01), which both were within the maximum allowable error range(±0.3 ℃) for disease diagnosis. Moreover, the MSE decreased from 0.38 to 0.05 (<0.01), while PSNR increased from 45.14 dB to 53.66 dB. In conclusion, the GMBI purposed in this study can filter the majority of noise caused by hair in temperature distribution and significantly improve skin temperature detection accuracy.
infrared imaging; temperature distribution; filter; pig; image interpolation; algorithms
賈桂鋒,蒙俊宇,武 墩,王登輝,高 云,馮耀澤. 被毛對(duì)熱成像檢測(cè)生豬體表溫度精度的影響及噪聲濾除方法[J]. 農(nóng)業(yè)工程學(xué)報(bào),2019,35(4):162-167. doi:10.11975/j.issn.1002-6819.2019.04.020 http://www.tcsae.org
Jia Guifeng, Meng Junyu, Wu Dun, Wang Denghui, Gao Yun, Feng Yaoze. Effect of hair on thermometry of skin by infrared thermography and noise reduction method for live pigs[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2019, 35(4): 162-167. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2019.04.020 http://www.tcsae.org
2018-10-12
2018-12-31
湖北省自然科學(xué)基金(2018CFB099);中央高校基本科研業(yè)務(wù)專(zhuān)項(xiàng)基金(2662016QD002);國(guó)家級(jí)大學(xué)生創(chuàng)新創(chuàng)業(yè)訓(xùn)練計(jì)劃(201810504076)
賈桂鋒,博士,講師,主要從事生豬智能化檢測(cè)技術(shù)與裝備。 Email:guifeng@mail.hzau.edu.cn
馮耀澤,博士,副教授,主要從事智能化檢測(cè)與控制技術(shù)。 Email:yaoze.feng@mail.hzau.edu.cn
10.11975/j.issn.1002-6819.2019.04.020
S818.2
A
1002-6819(2019)-04-0162-06