国产日韩欧美一区二区三区三州_亚洲少妇熟女av_久久久久亚洲av国产精品_波多野结衣网站一区二区_亚洲欧美色片在线91_国产亚洲精品精品国产优播av_日本一区二区三区波多野结衣 _久久国产av不卡

?

霧滴沉積特性參數(shù)的圖像檢測(cè)算法改進(jìn)

2018-09-03 01:42劉思瑤田素博李天來(lái)
關(guān)鍵詞:試紙圖像處理灰度

郭 娜,劉思瑤,須 暉,田素博,3,李天來(lái)

?

霧滴沉積特性參數(shù)的圖像檢測(cè)算法改進(jìn)

郭 娜1,2,3,劉思瑤2,須 暉1,3,田素博2,3,李天來(lái)1,3※

(1. 沈陽(yáng)農(nóng)業(yè)大學(xué)園藝學(xué)院,沈陽(yáng) 110866;2. 沈陽(yáng)農(nóng)業(yè)大學(xué)工程學(xué)院,沈陽(yáng) 110866; 3. 設(shè)施園藝省部共建教育部重點(diǎn)試驗(yàn)室,沈陽(yáng) 110866)

快速獲取施藥后霧滴在靶標(biāo)表面的沉積分布有助于了解農(nóng)藥的田間分布情況,水敏試紙霧滴圖像處理算法是檢測(cè)噴藥沉積特性參數(shù)常用的方法,但常因光照不均、試紙上沉積的霧滴斑痕粘連而引起霧滴識(shí)別誤差。為解決這一問(wèn)題,針對(duì)手機(jī)拍攝的照片,該文提出了與位置相關(guān)的動(dòng)態(tài)閾值法提取霧滴區(qū)域,并設(shè)計(jì)基于圓形度的循環(huán)分割算法對(duì)粘連霧滴分割。以水代替農(nóng)藥利用背負(fù)式噴霧器噴灑,選取8張不同稀疏程度和碰撞角度的試紙作為樣本進(jìn)行試驗(yàn),以驗(yàn)證上述算法的檢測(cè)效果。試驗(yàn)結(jié)果表明,該方法不受亮度不均影響,覆蓋率比固定閾值與分塊閾值法分別高12.57%和8.74%,提取到的霧滴區(qū)域更加完整,能夠提取92.64%以上的霧滴,且粘連分割效果較好,霧滴識(shí)別的正確率為97.2%,覆蓋密度檢測(cè)誤差為3.31%,能夠滿足實(shí)際生產(chǎn)要求,為下一步開(kāi)發(fā)霧滴檢測(cè)APP打下基礎(chǔ)。

圖像處理;圖像分割;算法;水敏試紙;霧滴;沉積分布;動(dòng)態(tài)閾值;粘連分割

0 引 言

設(shè)施蔬菜病蟲(chóng)害頻發(fā),化學(xué)防治仍然是主要手段,農(nóng)民為了追求高產(chǎn),農(nóng)藥濫用現(xiàn)象十分嚴(yán)重,導(dǎo)致農(nóng)產(chǎn)品農(nóng)藥殘留超標(biāo)、農(nóng)田環(huán)境污染等一系列問(wèn)題。如何提高農(nóng)藥利用率、減少農(nóng)藥使用量成為亟待解決的問(wèn)題。國(guó)內(nèi)外學(xué)者將傳感器技術(shù)、變量噴霧技術(shù)、航空技術(shù)、現(xiàn)代控制技術(shù)等先進(jìn)技術(shù)與施藥技術(shù)相結(jié)合以提高農(nóng)藥利用率,但如何準(zhǔn)確快速評(píng)價(jià)各種先進(jìn)施藥技術(shù)的噴施效果一直是田間測(cè)試中的難題。為了發(fā)揮農(nóng)藥的最佳效力達(dá)到有效防治的效果,霧化后的藥液霧滴必須均勻、適當(dāng)?shù)胤植荚谥参锶~片等靶標(biāo)上,并達(dá)到一定的富集量,霧滴在作物上的沉積分布與農(nóng)作物的病蟲(chóng)害防治效果密切相關(guān)。快速獲取施藥后霧滴田間沉積分布情況是目前衡量農(nóng)藥噴施效果的重要手段之一,可對(duì)施藥效果進(jìn)行定量評(píng)價(jià),為進(jìn)一步優(yōu)化農(nóng)藥噴施技術(shù)提供參考[1-3]。

目前霧滴沉積分布特性的田間檢測(cè)方法主要有直接檢測(cè)和間接檢測(cè)2種。直接檢測(cè)法利用熒光染色劑代替農(nóng)藥噴灑,然后用蒸餾水洗脫葉片上沉積的藥液,通過(guò)洗脫溶液的熒光劑濃度來(lái)得到霧滴的沉積量[4],該方法能夠直接得到作物葉片上的霧滴沉積量,但洗脫效果直接決定檢測(cè)結(jié)果,準(zhǔn)確性不高。間接檢測(cè)法是利用油盤(pán)、氧化鎂采樣板、水敏試紙等各種霧滴收集器代替葉片承接霧滴,再以人工統(tǒng)計(jì)或圖像處理的方法對(duì)收集器內(nèi)霧滴進(jìn)行分析統(tǒng)計(jì),具有直觀,適用性廣的優(yōu)點(diǎn)[5-12]。其中水敏試紙顯色明顯,易于圖像處理和保存,是目前最常用的霧滴收集器[13-15]。目前,國(guó)外已開(kāi)發(fā)了多種水敏試紙圖像處理軟件[16-21],國(guó)內(nèi)大疆公司也推出與智能移動(dòng)設(shè)備配合使用的霧滴分析設(shè)備和App,實(shí)現(xiàn)田間霧滴沉積分布的快速檢測(cè),但霧滴圖像識(shí)別效果有待進(jìn)一步提高。

水敏試紙上霧滴斑痕的圖像識(shí)別主要包括霧滴區(qū)域提取和粘連霧滴分割兩步。霧滴區(qū)域提取將圖像中顯色的霧滴區(qū)域與背景分割開(kāi),目前常用的算法有固定閾值法[22-24]、分塊閾值法[25],Ostu自動(dòng)閾值法[14,26],上述算法采用一個(gè)或多個(gè)固定閾值對(duì)圖像進(jìn)行分割,在圖像亮度不均的情況下提取效果較差。粘連霧滴分割將試紙上部分重疊的霧滴進(jìn)行分割,常用的算法有基于分水嶺分割法[27]和尋找分離點(diǎn)對(duì)的算法,前者對(duì)微弱邊緣十分敏感,圖像中的噪聲、細(xì)微的灰度變化,都有可能產(chǎn)生過(guò)度分割的現(xiàn)象,導(dǎo)致霧滴圖像信息的丟失;后者不適于許多霧滴粘連在一起的情況[28]。

為提高噴藥沉積特性檢測(cè)的精度與效率,本文對(duì)水敏試紙上霧滴的圖像處理算法進(jìn)行了優(yōu)化改進(jìn):利用動(dòng)態(tài)閾值算法解決圖像亮度不均對(duì)霧滴區(qū)域提取的影響;采用基于圓形度的循環(huán)判斷分割法,解決粘連霧滴區(qū)域不易分割或過(guò)度分割的問(wèn)題,實(shí)現(xiàn)對(duì)霧滴沉積性能參數(shù)的檢測(cè),并試驗(yàn)驗(yàn)證本文圖像處理算法的檢測(cè)效果。

1 材料與方法

1.1 水敏試紙圖像采集

水敏試紙本色為黃色,遇水區(qū)域迅速變?yōu)樗{(lán)綠色,反應(yīng)迅速,對(duì)比明顯,易于圖像處理。本文采用普蘭迪機(jī)電設(shè)備公司生產(chǎn)的背負(fù)式電動(dòng)噴霧機(jī)(工作壓力0.2~0.4 MPa,額定電壓12 V)及杭州美琪公司生產(chǎn)的L334單孔錐形噴頭搭建噴霧系統(tǒng),噴頭霧化角為67°,試驗(yàn)獲得的水敏試紙放在自制的圖像采集裝置上,智能手機(jī)放到水平支架上,并保持水敏試紙平行,如圖1所示,在自然光照條件下拍攝水敏試紙圖像,然后上傳到電腦上進(jìn)行圖像處理。圖像處理程序采用德國(guó)MVTec公司研發(fā)的HALCON專(zhuān)業(yè)圖像處理軟件進(jìn)行編制,可提供豐富的函數(shù)庫(kù),分析方便精準(zhǔn),應(yīng)用廣泛[29-30]。

1.智能手機(jī) 2.水敏試紙 3.標(biāo)定板 4.支架

1.2 霧滴沉積特征參數(shù)表示方法

利用水敏試紙法測(cè)量霧滴沉積特性時(shí),采用霧滴覆蓋率和霧滴密度來(lái)衡量采樣點(diǎn)的霧滴沉積情況,進(jìn)而以各采樣點(diǎn)的變異系數(shù)作為霧滴田間分布均勻性的指標(biāo)。

1.2.1 霧滴覆蓋率

霧滴覆蓋率(coverage percentage)用霧滴覆蓋區(qū)域面積占統(tǒng)計(jì)總面積的百分比表示,如公式(1)所示。

式中為噴霧覆蓋率,%;A為霧滴區(qū)域像素?cái)?shù);A為試紙區(qū)域總像素?cái)?shù)。

1.2.2 霧滴覆蓋密度

霧滴覆蓋密度(coverage density)是影響霧滴防治效果的重要參數(shù),是指霧滴收集裝置單位面積上所承接的霧滴個(gè)數(shù),如公式(2)所示。

式中為霧滴覆蓋密度,個(gè)/cm2;為霧滴的總數(shù);為試紙的總面積,cm2。

1.3 水敏試紙圖像處理算法

水敏試紙圖像經(jīng)過(guò)圖像預(yù)處理、霧滴區(qū)域提取、粘連霧滴分割后即可得到霧滴沉積特征參數(shù)。

1.3.1 圖像預(yù)處理

本文對(duì)水敏試紙圖像進(jìn)行傅里葉變換增強(qiáng)后,利用標(biāo)定板校正鏡頭畸變,確定物理尺寸和像素間的換算關(guān)系,避免拍攝角度等因素的影響,提高圖像檢測(cè)精度,然后將彩色圖像分為R-G-B三通道圖像,如圖2所示。其中R通道試紙區(qū)域與背景的對(duì)比度更強(qiáng),B通道圖像除試紙區(qū)域外灰度值與R通道相近,因此本文選取R通道圖像與B通道圖像做差,得到僅有水敏試紙區(qū)域與背景對(duì)比非常明顯的灰度圖像,利用灰度的明顯差異分割出水敏試紙,并計(jì)算水敏試紙區(qū)域?qū)嶋H面積。

圖2 水敏試紙圖像預(yù)處理

1.3.2 霧滴區(qū)域提取

霧滴區(qū)域提取是將水敏試紙的灰度圖像進(jìn)行分割,將顯色的霧滴區(qū)域圖像與背景分割開(kāi)。閾值法是最常用的快速分割算法,設(shè)圖像中某一像素點(diǎn)的灰度值為(,),根據(jù)一定的方法選取某個(gè)灰度值作為閾值,根據(jù)公式 (3)進(jìn)行對(duì)圖像進(jìn)行灰度變換,變換后的圖像灰度為(,),對(duì)于霧滴圖像元素灰度為0(黑),對(duì)于背景的圖像元素灰度為255(白)。

選擇正確的閾值是分割成功的關(guān)鍵。目前水敏試紙圖像處理中最常用的閾值化算法為固定閾值法、Ostu自動(dòng)閾值法和分塊閾值法。前兩種方法利用圖像灰度直方圖或類(lèi)間方差準(zhǔn)則標(biāo)準(zhǔn)選取分割閾值,后一種先將圖像分成幾塊,根據(jù)每塊圖像的實(shí)際情況選擇分割閾值。本文利用智能手機(jī)在自然光條件下采集水敏試紙圖像,圖像亮度均勻性較差,較亮處?kù)F滴的灰度值甚至高于較暗處背景的灰度值,利用一個(gè)或幾個(gè)閾值進(jìn)行霧滴區(qū)域提取的效果較差。

本文采用動(dòng)態(tài)閾值法進(jìn)行霧滴區(qū)域提取。動(dòng)態(tài)閾值法選用與像素位置相關(guān)的一組閾值(,)來(lái)對(duì)圖像進(jìn)行分割。在水敏試紙的灰度圖像中,霧滴區(qū)域的灰度值遠(yuǎn)低于其鄰域背景的灰度值,因此本文以待分割點(diǎn)(,)為中心,矩形區(qū)域×上所有像素的平均灰度值作為該點(diǎn)的分割閾值,如公式(4)所示。

由于采取的均值濾波的結(jié)構(gòu)元素過(guò)小會(huì)使得霧滴區(qū)域提取不全面,過(guò)大會(huì)使得霧滴與背景對(duì)比信息損失,本文經(jīng)過(guò)多次試驗(yàn)調(diào)整,最終確定以25×25 pixel的矩形作為結(jié)構(gòu)元素,對(duì)待分割的水敏試紙灰度圖像進(jìn)行均值濾波處理,得到待分割圖像每個(gè)像素點(diǎn)的閾值,并構(gòu)成閾值圖像。比較待分割圖像和閾值圖像的每個(gè)像素點(diǎn)的灰度值,會(huì)發(fā)現(xiàn)霧滴區(qū)域邊緣的像素灰度值會(huì)增大,而背景區(qū)域某些的像素灰度值也會(huì)稍有變化,為防止這些灰度值增加的背景像素對(duì)提取精度的影響,因此需要對(duì)處理后的圖像進(jìn)行灰度值補(bǔ)償Offset(本文選取-15),然后根據(jù)公式(5)即可對(duì)水敏試紙圖像進(jìn)行二值化處理

利用動(dòng)態(tài)閾值法對(duì)霧滴區(qū)域的分割過(guò)程如圖3所示。分割后即可根據(jù)公式(1)求取霧滴覆蓋率。

圖3 動(dòng)態(tài)閾值法對(duì)霧滴區(qū)域的分割過(guò)程

1.3.3 粘連霧滴分割

霧滴撞擊水敏試紙時(shí)經(jīng)常出現(xiàn)2個(gè)或多個(gè)霧滴落在相近位置,造成霧滴斑痕重疊的現(xiàn)象,因此欲得到水敏試紙上準(zhǔn)確的霧滴個(gè)數(shù),需將粘連的霧滴區(qū)域進(jìn)行分割。這些粘連的霧滴有許多形狀特性與單獨(dú)的霧滴不同,如圓形度。一個(gè)區(qū)域的圓形度表示該區(qū)域與圓形的相似度,其計(jì)算方法如公式(6)所示。

式中為區(qū)域的面積,mm2;max為區(qū)域邊界點(diǎn)距中心點(diǎn)的最大距離,mm;

圓形度0.5常被作為判斷霧滴粘連的閾值[18],本文對(duì)試紙上肉眼容易辨別的獨(dú)立霧滴與粘連霧滴各500個(gè)進(jìn)行圓形度統(tǒng)計(jì)分析,結(jié)果表明獨(dú)立霧滴平均圓形度為0.75,粘連霧滴平均圓形度為0.41,圓形度差異較大,但部分粘連嚴(yán)重的霧滴區(qū)域圓形度接近于0.5,因此本文選定圓形度0.6作為區(qū)分粘連霧滴與單獨(dú)霧滴的閾值,然后對(duì)粘連霧滴區(qū)域進(jìn)行先腐蝕再膨脹的結(jié)構(gòu)化分割操作,通過(guò)腐蝕操作將粘連霧滴區(qū)域分割開(kāi),并進(jìn)行連通域處理和標(biāo)記,然后進(jìn)行膨脹操作避免了霧滴區(qū)域大小的改變。

粘連霧滴區(qū)域分割中發(fā)現(xiàn)單次分割對(duì)于2個(gè)以上霧滴粘連區(qū)域效果不好,因此本文設(shè)計(jì)了如圖4所示的循環(huán)判斷分割算法對(duì)試紙上粘連霧滴進(jìn)行多次分割。首先將二值化的霧滴圖像進(jìn)行連通域處理和標(biāo)記,然后選取圓形度小于0.6的區(qū)域,作為待分割區(qū)域,選取半徑為(0=1 pixel)的圓形為結(jié)構(gòu)元素,對(duì)粘連區(qū)域進(jìn)行一次先腐蝕后膨脹的結(jié)構(gòu)化分割操作,接著對(duì)分割后的區(qū)域重新進(jìn)行連通域處理及圓形度計(jì)算,再次提取分割后區(qū)域中圓形度小于0.6的區(qū)域,并增加腐蝕和膨脹的系數(shù)=+1,進(jìn)入下一輪分割操作,直至所有霧滴區(qū)域的圓形度均大于0.6,則認(rèn)為粘連的霧滴區(qū)域全部分割完成。分割過(guò)程中對(duì)圓形度大于0.6的區(qū)域不進(jìn)行處理,防止細(xì)小獨(dú)立霧滴被腐蝕后丟失。經(jīng)過(guò)循環(huán)分割后,統(tǒng)計(jì)圖像中所有分割后互不相連的霧滴區(qū)域個(gè)數(shù),然后再除以水敏試紙圖像面積,即可得到霧滴密度。

注:r為腐蝕膨脹系數(shù)。

2 試驗(yàn)驗(yàn)證

本文將背負(fù)式噴霧器噴桿水平放置,以水代替農(nóng)藥,從水敏試紙(30 cm×35 cm)上方50 cm向下噴灑,并調(diào)整試紙距離噴出霧錐中心軸的距離,使霧滴約以角度撞擊到試紙上,從中挑選8張?jiān)嚰堊鳛闃颖具M(jìn)行分析,依次進(jìn)行霧滴覆蓋率和覆蓋密度試驗(yàn),以驗(yàn)證本文算法的檢測(cè)效果。根據(jù)人工計(jì)數(shù)得到的霧滴密度大小,分為稀疏、適中、密集3組[21],如圖5所示。其中,稀疏組霧滴密度<80,碰撞角度=90°;適中組霧滴密度80<<110,撞擊角度依次為=70°、50°、30°、15°;密集組霧滴密度>110,碰撞角度=90°。

2.1 霧滴覆蓋率檢測(cè)

2.1.1 霧滴區(qū)域提取算法對(duì)比試驗(yàn)

在自然光照下采集稀疏2試紙的圖像,如圖6所示,因光照不均造成圖像左暗右亮,分別應(yīng)用固定閾值法、分塊閾值法以及動(dòng)態(tài)閾值算法對(duì)其進(jìn)行霧滴區(qū)域提取,對(duì)比不同霧滴區(qū)域提取算法對(duì)亮度不均圖像的處理效果,前兩種方法的分割閾值按照參考文獻(xiàn)[14]確定,結(jié)果如圖6所示。

注:K為霧滴覆蓋密度,個(gè)·cm-2;α為霧滴撞擊試紙角度,(°)。

圖6 霧滴區(qū)域提取方法對(duì)比試驗(yàn)結(jié)果

對(duì)比圖6b和6a可知,左側(cè)提取的霧滴區(qū)域明顯比原圖大,造成許多本未粘連的霧滴反而粘連在一起,而右側(cè)提取的霧滴區(qū)域則明顯小于原圖,其原因是分割閾值對(duì)于左側(cè)暗區(qū)較大,部分較暗試紙被提取為霧滴,同理該閾值對(duì)右側(cè)亮區(qū)較小,較淺霧滴區(qū)域未被提取。由圖6c可知,試紙圖像被分成4個(gè)區(qū)域,每個(gè)區(qū)域選用不同的分割閾值,與固定閾值相比,左側(cè)霧滴區(qū)域明顯減小,右側(cè)霧滴區(qū)域明顯增加,且與原圖中霧滴大小更接近,但仍有部分較淺霧滴斑痕無(wú)法準(zhǔn)確提取,如圖6c中方框所示;由圖6d可知,圖像亮度不均勻?qū)Ρ疚奶岢龅膭?dòng)態(tài)閾值法影響并不大,試紙上顯色的霧滴均被準(zhǔn)確提取出來(lái),與分塊閾值相比,左上部細(xì)小霧滴也被提取出來(lái),且提取的霧滴區(qū)域面積更接近于原圖。

2.1.2 霧滴覆蓋率對(duì)比試驗(yàn)

分別應(yīng)用固定閾值法、分塊閾值法以及本文所述動(dòng)態(tài)閾值算法求取8張樣本的霧滴覆蓋密度,結(jié)果如表1所示。

表1 霧滴覆蓋率檢測(cè)結(jié)果

根據(jù)公式(7)計(jì)算8張圖像在3種閾值分割方法下覆蓋率的平均相對(duì)誤差。固定閾值法和分塊閾值法與本文算法的霧滴覆蓋率平均相對(duì)誤差分別為12.57%和8.74%,其中密集2試紙的效果最明顯,霧滴覆蓋率提高了19.6%和15.4%。

人工統(tǒng)計(jì)各個(gè)試紙上未提取的霧滴個(gè)數(shù),按照?qǐng)D5中試紙的順序依次為5、9、0、55、8、65、117、138個(gè),占霧滴總個(gè)數(shù)的0.96%、1.21%、0、6.48%、1.04%、7.36%、5.13%、4.57%,平均相對(duì)誤差為3.35%,因此本文霧滴提取算法能夠提取試紙上92.64%以上的霧滴。由圖5可知,對(duì)于霧滴較清晰的稀疏1、稀疏2、適中1、適中3試紙,未提取的霧滴個(gè)數(shù)較少,其他試紙因存在部分顏色極淺細(xì)小霧滴,霧滴提取難度大,未提取霧滴個(gè)數(shù)較多。

因此,本文算法提取的霧滴區(qū)域更加完整,不受圖像亮度不均的影響,且對(duì)于較淺的細(xì)小霧滴提取效果有所提高。

2.2 霧滴覆蓋密度檢測(cè)

2.2.1 粘連霧滴分割試驗(yàn)

本試驗(yàn)以圖5f霧滴覆蓋密度=102.67個(gè)/cm2,撞擊角度=15°的試紙為例,直觀體現(xiàn)本文所述算法對(duì)粘連霧滴的分割效果,識(shí)別出的霧滴以空心圓表示,并將4個(gè)識(shí)別錯(cuò)誤粘連霧滴以方框標(biāo)記,如圖7所示。

注:1為過(guò)度分割霧滴,2為未正確分割霧滴。

由圖7可知,粘連霧滴循環(huán)分割算法能夠準(zhǔn)確分割大部分粘連霧滴,尤其是多個(gè)霧滴粘連的情況,但仍有部分粘連霧滴被過(guò)度分割或未識(shí)別。由于霧滴與試紙f的碰撞角度較小,大霧滴斑痕均細(xì)長(zhǎng)且拖尾,圖中部分狹長(zhǎng)霧滴因圓形度過(guò)小而被過(guò)度分割,如方框1中單個(gè)細(xì)長(zhǎng)霧滴被分割為3個(gè)霧滴;另外部分與大霧滴相連的細(xì)小霧滴也未被準(zhǔn)確分割,如圖中方框2中左側(cè)凸起的小霧滴,其原因是霧滴粒徑相差較大,小霧滴對(duì)整個(gè)連通區(qū)域的圓形度影響較小,因此該區(qū)域被劃分為獨(dú)立霧滴而未被分割。

人工統(tǒng)計(jì)8張?jiān)嚰堉幸蜻^(guò)度分割而增加的霧滴以及未準(zhǔn)確分割而丟失的霧滴總個(gè)數(shù),按照?qǐng)D5中試紙的順序依次為3、8、17、5、33、72、80、53個(gè),占霧滴個(gè)數(shù)的0.57%、1.07%、2.3%、0.58%、4.29%、8.16%、3.51%、1.76%,霧滴識(shí)別的正確率為97.2%。稀疏1和適中2試紙霧滴粘連程度較輕,本文所設(shè)計(jì)的霧滴循環(huán)分割算法能夠準(zhǔn)確分割粘連霧滴,霧滴識(shí)別誤差較小。圖7中試紙因霧滴細(xì)長(zhǎng)造成分割誤差較多,因此水敏試紙應(yīng)用時(shí)應(yīng)水平放置,避免霧滴以小角度碰撞試紙。

各個(gè)試紙粘連霧滴循環(huán)分割時(shí)間依次為89、108、151、138、153、180、176、238 ms,分割時(shí)間隨霧滴個(gè)數(shù)和粘連程度的增加而增加,但總體耗時(shí)較少。因此,本文的霧滴循環(huán)分割算法能夠有效分割2個(gè)尤其多個(gè)霧滴粘連的情況,對(duì)于少數(shù)細(xì)長(zhǎng)霧滴和粘連嚴(yán)重且形狀不規(guī)則的霧滴分割效果較差,需進(jìn)一步優(yōu)化改進(jìn)。

2.2.2 霧滴覆蓋密度對(duì)比試驗(yàn)

本文分別計(jì)算了8張?jiān)嚰垬颖旧系母采w密度,將其與人工計(jì)數(shù)法檢測(cè)的結(jié)果進(jìn)行對(duì)比分析,結(jié)果如表2所示。

表2 霧滴覆蓋密度檢測(cè)結(jié)果

本文分割算法得到的霧滴覆蓋密度,與人工計(jì)算方法結(jié)果相比,平均相對(duì)誤差為3.31%,最大相對(duì)誤差為6.78%,表明本文的粘連霧滴分割算法能夠有效分割人眼所能識(shí)別的大部分粘連霧滴,滿足實(shí)際應(yīng)用的要求。

3 結(jié) 論

為實(shí)現(xiàn)噴藥沉積特性參數(shù)的快速檢測(cè),本文利用日益強(qiáng)大的智能手機(jī)拍照功能代替專(zhuān)用的圖像采集系統(tǒng),對(duì)水敏試紙的圖像處理算法進(jìn)行了改進(jìn),并進(jìn)行了試驗(yàn)驗(yàn)證,主要研究?jī)?nèi)容包括:

1)針對(duì)霧滴覆蓋檢測(cè)時(shí),光照條件不均導(dǎo)致的霧滴區(qū)域提取困難問(wèn)題,提出了基于動(dòng)態(tài)閾值的水敏試紙圖像二值化方法,提取到的霧滴區(qū)域信息完整。試驗(yàn)證明本文算法不受亮度不均影響,覆蓋率檢測(cè)結(jié)果分別比固定閾值法與分塊閾值法高12.57%和8.74%,且能夠提取試紙上92.64%以上的霧滴。

2)針對(duì)霧滴覆蓋密度檢測(cè)時(shí)粘連霧滴區(qū)域?qū)F滴個(gè)數(shù)統(tǒng)計(jì)的干擾,本文提出了基于圓形度的循環(huán)分割法對(duì)粘連霧滴區(qū)域進(jìn)行分割,本文的算法能夠有效分割大部分粘連嚴(yán)重霧滴和細(xì)長(zhǎng)霧滴,霧滴識(shí)別的正確率為97.2%,覆蓋密度與人工計(jì)數(shù)誤差為3.31%,對(duì)于細(xì)長(zhǎng)霧滴和粘連嚴(yán)重霧滴的分割可進(jìn)一步優(yōu)化算法。

綜上所述,本文所設(shè)計(jì)的霧滴沉積特性參數(shù)圖像檢測(cè)算法可以滿足實(shí)際生產(chǎn)中對(duì)霧滴覆蓋率及覆蓋密度檢測(cè)的需求。為實(shí)現(xiàn)在實(shí)際生產(chǎn)條件下快速準(zhǔn)確地檢測(cè),還需進(jìn)一步對(duì)圖像處理算法進(jìn)行改進(jìn)以適應(yīng)環(huán)境的干擾,并開(kāi)發(fā)配套的手機(jī)APP。

[1] Holownicki R, Doruchowski G, Godyn A, et al. Variation of sprav deposit and loss with air-jet directions applied in orchards[J]. Journal of Agricultural Engineering Research, 2000, 77(2): 129-136.

[2] Murphy S D, Miller P C H, Parkin C S. The Effect of boom section and nozzle configuration on the risk of spray drift[J]. Journal of Agricultural Engineering Research, 2000, 75: 127-137.

[3] Wise J C, Jenkins P E, Schilder A M C, et al. Sprayer type and water volume influence pesticide deposition and control of insect pests and diseases in juice grapes[J]. Crop Protection, 2010, 29: 378-385.

[4] 陳菁菁,彭彥昆,李永玉,等. 基于高光譜熒光技術(shù)的葉菜農(nóng)藥殘留快速檢測(cè)[J].農(nóng)業(yè)工程學(xué)報(bào),2010,26(增刊2):1-5.

Chen Jingjing, Peng Yankun, Li Yongyu, et al. Rapid detection of vegetable pesticide residue based on hypersectral fluorescence imaging technology[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(Supp.2): 1-5. (in Chinese with English abstract)

[5] 馮斌,汪懋華. 基于顏色分形的水果計(jì)算機(jī)視覺(jué)分級(jí)技術(shù)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2002,18(2):141-144.

Feng Bin, Wang Maohua. Computer vision classification of fruit based on fractal color[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2002, 18(2): 141-144. (in Chinese with English abstract)

[6] 陳怡群,常春,湯伯敏. 霧滴顯微圖像分析中處理亮度不均勻性的算法[J]. 農(nóng)業(yè)工程學(xué)報(bào),2009,25(1):149-153.

Chen Yiqun, Chang Chun, Tang Bomin. Algorithm for handling uneven brightness in image processing of fogdrop micrograph[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2009, 25(1): 149-153. (in Chinese with English abstract)

[7] 韓仲志,趙友剛,楊錦忠. 基于籽粒RGB圖像獨(dú)立分量的玉米胚部特征檢測(cè)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2010,26(3):222-226.

Han Zhongzhi, Zhao Yougang, Yang Jinzhong. Detection of embryo based on independent components for kernel RGB images in maize[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2010, 26(3): 222-226. (in Chinese with English Abstract)

[8] 張京,宋堅(jiān)利,何雄奎,等. 扇形霧噴頭霧化過(guò)程中霧滴運(yùn)動(dòng)特性[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2011,42(4):66-69. Zhang Jing, Song Jianli, He Xiongkui, et al. Droplets movement characteristics in atomization process of flat fan nozzle[J]. Transactions of The Chinese Society for Agricultural Machinery, 2011, 42(4): 66-69. (in Chinese with English Abstract)

[9] 宋淑然,王衛(wèi)星,洪添勝,等. 水稻田農(nóng)藥噴霧上層植株霧滴截留影響的試驗(yàn)研究[J]. 農(nóng)業(yè)工程學(xué)報(bào),2003,19(6):115-117.

Song Shuran, Wang Weixing, Hong Tiansheng, et al. Testing research on effects of top layer rice fog drop interception on pesticide spraying distribution in rice fields [J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2003, 19(6): 115-117. (in Chinese with English Abstract)

[10] 楊洲,牛萌萌,李君,等. 果園在線混藥型靜電噴霧機(jī)的設(shè)計(jì)與試驗(yàn)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(21):60-67.

Yang Zhou, Niu Mengmeng, Li Jun, et al. Design and experiment of an electrostatic sprayer with on-line mixing system for orchard[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(21): 60-67. (in Chinese with English abstract)

[11] 陳益民,張性雄,何金成,等. 微霧滴取樣與數(shù)據(jù)采集方法[J]. 福建農(nóng)林大學(xué)學(xué)報(bào):自然科學(xué)版,2006,35(4):445-448.

Chen Yimin, Zhang Xingxiong, He Jincheng, et al. Sprayed droplet sampling and data gathering[J]. Journal of Fujian Agriculture and Forestry University: Natural Science Edition, 2006, 35(4): 445-448. (in Chinese with English abstract)

[12] 李彥,史慧勤,楊振洲,等. 衛(wèi)生殺蟲(chóng)劑霧滴直徑測(cè)定方法的比較研究[J]. 中華衛(wèi)生殺蟲(chóng)藥械,2014,20(1):34-36.

Li Yan. Shi Huiqin, Yang Zhenzhou, et al. A comparative study on methods for testing the size of health insecticide droplet[J]. Chinese Journal of Hygienic Insecticides & Equipments, 2014, 20(1): 34-36. (in Chinese with English abstract)

[13] 茹煜,金蘭,賈志成,等. 無(wú)人機(jī)靜電噴霧系統(tǒng)設(shè)計(jì)及試驗(yàn)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(8):42-47.

Ru Yu, Jin Lan, Jia Zhicheng, et al. Design and experiment on electrostatic spraying system for unmanned aerial vehicle[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(8): 42-47. (in Chinese with English abstract)

[14] 蔣煥煜,張利君,劉光遠(yuǎn),等. 基于PWM變量噴霧的單噴頭動(dòng)態(tài)霧量分布均勻性試驗(yàn)[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2017,48(4):41-46.

Jiang Huanyu, Zhang Lijun, Liu Guangyuan, et al. Experiment on dynamic spray deposition uniformity for PWM variable spray of single nozzle[J]. Journal of Agricultural Machinery, 2017, 48(4): 41-46. (in Chinese with English Abstract)

[15] 鄭琦琦,吳堅(jiān). 基于NO2-顯色反應(yīng)的抗?jié)裥試婌F霧滴密度和大小的檢測(cè)體系[J]. 農(nóng)業(yè)工程學(xué)報(bào),2015,31(3):107-112.

Zheng Qiqi, Wu Jian. Humidity resistance test system for spray droplet density and size based on NO2-color reaction[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2015, 31(3): 107-112. (in Chinese with English abstract)

[16] Nansen C, Ferguson J C, Moore J, et al. Optimizing pesticide spray coverage using a novel web and smartphone tool, SnapCard[J]. Agronomy for Sustainable Development, 2015, 35(3): 1075-1085.

[17] Salyani M, Serdynski J. Development of a sensor for spray deposition assessment[J]. Transactions of the ASAE, 1990, 90: 1464-1469.

[18] Cunha J P A R, Farnese A C, Olivet J J. Computer programs for analysis of droplets sprayed on water sensitive papers[J]. Planta Daninha, 2013, 31(3): 715-720.

[19] Ferguson J C, Hewitt A J, Donnell C C. Pressure, droplet size classification, and nozzle arrangement effects on coverage and droplet number density using air-inclusion dual fan nozzles for pesticide applications[J]. Crop Protection, 2016, 89: 231-238.

[20] Ferguson J C, Chechetto R G, O’Donnell C C, et al. Assessing a novel smart phone application-SnapCard, compared to five imaging systems to quantify droplet deposition on artificial collectors[J]. Computers and Electronics in Agriculture, 2016, 728: 793-798.

[21] Cunha M, Carvalho C, Ars M. Assessing the ability of image processing software to analyse spray quality on water-sensitive papers used as artificial targets[J]. Biosystems Engineering, 2012, 111(1): 11-23.

[22] 杜慧茜,梅文博,蘇廣川. 一種固定閾值的毫米波圖像邊緣檢測(cè)方法[J]. 微電子學(xué)與計(jì)算機(jī),2005,22(4):1-3,7.

Du Huiqian, Mei Wenbo, Su Guangchuan. A fixed threshold edge detection approach for millimeter wave images[J]. Microelectronics & Computer, 2005, 22(4): 1-3,7. (in Chinese with English Abstract)

[23] 段晉英,史建芳. 改進(jìn)的高低帽變換對(duì)固定閾值二值化算法的優(yōu)化[J]. 科學(xué)技術(shù)與工程,2014,14(15):245-249.

Duan Jinying, Shi Jianfang. The optimization of binary conversion with a definite value by the improved top-hat and bottom-hat transformation[J]. Science Technology and Engineering, 2014, 14(15): 245-249. (in Chinese with English Abstract)

[24] 祁力鈞,胡開(kāi)群,莽璐,等. 基于圖像處理的霧滴檢測(cè)技術(shù)[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2009,40(增刊1):48-51. Qi Lijun, Hu Kaiqun, Mang Lu, et al. Droplet detection based on image processing[J]. Journal of Agricultural Machinery, 2009, 40(Supp.1): 48-51. (in Chinese with English Abstract)

[25] 袁江濤,楊立,王小川,等. 基于機(jī)器視覺(jué)的細(xì)水霧液滴尺寸測(cè)量與分析[J]. 光學(xué)學(xué)報(bào),2009,29(10):2842-2847.

Yuan Jiangtao, Yang Li, Wang Xiaochuan, et al. Measurement and analysis of water mist droplet size based on machine vision[J]. Acta Optica Sinica, 2009, 29(10): 2842-2847. (in Chinese with English Abstract)

[26] 邱白晶,沙俊炎,湯伯敏,等. 密閉空間霧滴沉積狀態(tài)參數(shù)的顯微圖像解析[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2008,39(2):55-58.

Qiu Baijing, Sha Junyan, Tang Bomin, et al. Analysis of droplets deposition parameters based on micrograph in airtight space[J]. Transactions of the Chinese Society for Agricultural Machinery, 2008, 39(2): 55-58. (in Chinese with English Abstract)

[27] 梁萍,程偉,基于分水嶺的粘連霧滴圖像分割方法[J]. 重慶理工大學(xué)學(xué)報(bào):自然科學(xué)版,2008,22(10):170-174.

Liang Ping, Cheng Wei. Research of conglutinated droplets image segmentation based on watershed algorithm[J]. Journal of Chongqing Institute of Technology:Natural science, 2008, 22(10): 170-174. (in Chinese with English Abstract)

[28] Guimaraes L V, Suzim A A, Maeda J. A new automatic circular decomposition algorithm applied to blood cells image[C]//Proceedings of IEEE International Symposium on Bio-Informatics and Biomedical, 2000: 277-280.

[29] 黃亦其,黃體森,黃媚章,等. 基于局部均值的甘蔗莖節(jié)識(shí)別[J]. 中國(guó)農(nóng)機(jī)化學(xué)報(bào),2017,38(2):76-80.

Huang Yiqi, Huang Tisen, Huang Meizhang, et al. Recognition of sugarcane based on local mean[J]. Journal of Chinese Agricultural Mechanization, 2017, 38(2): 76-80. (in Chinese with English Abstract)

[30] 羅陸鋒,鄒湘軍,葉敏,等. 基于雙目立體視覺(jué)的葡萄采摘防碰空間包圍體求解與定位[J]. 農(nóng)業(yè)工程學(xué)報(bào),2016,32(8):41-47.

Luo Lufeng, Zou Xiangjun, Ye Min, et al. Calculation and localization of bounding volume of grape for undamaged fruit picking based on binocular stereo vision[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2016, 32(8): 41-47. (in Chinese with English Abstract)

Improvement on image detection algorithm of droplets deposition characteristics

Guo Na1,2,3, Liu Siyao2, Xu Hui1,3, Tian Subo2,3, Li Tianlai1,3※

(1.,,110866,; 2.,,110866,; 3.,110866,)

Droplets deposition characteristics estimation is helpful to know the pesticides deposition distribution on crops, which is related to the crop spray quality directly, and especially the fast detection method will provide the basis for improving the pesticides spraying technology. Droplets image processing based on the water sensitive paper is one of the most common methods to detect the droplets deposition characteristics. The droplets coverage percentage and coverage density were selected to evaluate deposition distribution in this paper. With the development of science and technology, a smart phone was selected as image acquiring tool to replace the special image acquisition system, and the improved image processing algorithm of the water sensitive paper was developed. There are 3 steps in the image processing algorithm to obtain each droplet stain, which are image preprocessing, droplets area segmentation, and overlapped droplets segmentation. Firstly, the image was enhanced and calibrated by a calibration board, and then the image of water sensitive paper was segmented from the R channel and B channel image and transferred to a gray scale image. Secondly, the blue droplets stain area was segmented from yellow paper background, and the dynamic threshold method based on the pixel position was used to solve the problem of the influence of uneven brightness in this step, in which the gray mean value of an area of 25×25 pixels was calculated as the segmentation threshold for the middle pixel, and the droplets coverage percentage was calculated by the stains pixel number divided by the pixel number of water sensitive paper area. Thirdly, the circulatory segmentation method based on region circularity was designed to segment the multiple-droplet overlapped regions.Based on the statistical analysis, 0.6 was selected as the circularity threshold, less than 0.6 was considered to be overlapped droplets, and first erosion and then dilation based on a coefficient of corrosion and expansionwas performed; next the segmented area circularity was recalculated, the area with the circularity of less than 0.6 was selected again and segmented by the coefficient (+1), and the erosion-dilation operation would be repeated over and over until the circularity of all stains was greater than 0.6. Finally, the identified droplets were marked as circle, and the droplets coverage density was calculated by droplets number divided by the water sensitive paper area. Experiments were conducted to test and verify the detection advancement of the proposed image processing algorithm. Experimental results showed that the dynamic threshold segmentation method is not affected by the uneven brightness and can extract 92.64% of droplets, and the droplet coverage percentage detection result is 12.57% and 8.74% greater than constant threshold and partitioned threshold method respectively. Moreover, the proposed overlapped droplets segmentation algorithm can segment successfully more than 2 droplets and the long and thin droplets, the accuracy of droplets identification is 97.2%, and the coverage density detection results showed that the relative error between the algorithm in this paper and manual counting is only 3.31%. The results indicated that the proposed image detection algorithm of droplets deposition characteristics is efficient and convenient, and can completely fulfill the demand of droplets deposition characteristics detection in the field, and the corresponding smartphone applications are in development.

image processing; image segmentation; algorithm; water sensitive paper; droplets; spray deposition; dynamic threshold; overlapped region segmentation

10.11975/j.issn.1002-6819.2018.17.023

TP391.41

A

1002-6819(2018)-17-0176-07

2018-03-02

2018-05-23

國(guó)家重點(diǎn)研發(fā)計(jì)劃(2016YFD0200708)

郭 娜,講師,博士,研究方向:設(shè)施園藝智能裝備的研究。Email:guona_stacy@163.com

李天來(lái),院士,博士生導(dǎo)師,主要從事蔬菜設(shè)施栽培研究。Email:ltl@syau.edu.cn

郭 娜,劉思瑤,須 暉,田素博,李天來(lái). 霧滴沉積特性參數(shù)的圖像檢測(cè)算法改進(jìn)[J]. 農(nóng)業(yè)工程學(xué)報(bào),2018,34(17):176-182. doi:10.11975/j.issn.1002-6819.2018.17.023 http://www.tcsae.org

Guo Na, Liu Siyao, Xu Hui, Tian Subo, Li Tianlai. Improvement on image detection algorithm of droplets deposition characteristics[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(17): 176-182. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2018.17.023 http://www.tcsae.org

猜你喜歡
試紙圖像處理灰度
采用改進(jìn)導(dǎo)重法的拓?fù)浣Y(jié)構(gòu)灰度單元過(guò)濾技術(shù)
海戰(zhàn)場(chǎng)偵察圖像處理技術(shù)圖譜及應(yīng)用展望
人工智能輔助冠狀動(dòng)脈CTA圖像處理和診斷的研究進(jìn)展
Bp-MRI灰度直方圖在鑒別移行帶前列腺癌與良性前列腺增生中的應(yīng)用價(jià)值
血糖試紙保存方法及注意事項(xiàng)
Arduino小車(chē)巡線程序的灰度閾值優(yōu)化方案
基于ARM嵌入式的關(guān)于圖像處理的交通信號(hào)燈識(shí)別
機(jī)器學(xué)習(xí)在圖像處理中的應(yīng)用
血糖試紙不要放進(jìn)冰箱