王懷宇 李景麗
摘要:對(duì)玉米常見(jiàn)雜草進(jìn)行灰度化、圖像濾波等圖像預(yù)處理,對(duì)多個(gè)圖像紋理特征進(jìn)行篩選。以支持向量機(jī)進(jìn)行分類(lèi)識(shí)別,分別進(jìn)行基于灰度矩陣、統(tǒng)計(jì)矩的識(shí)別以及二者結(jié)合的識(shí)別。結(jié)果表明,綜合了灰度矩陣與統(tǒng)計(jì)矩的紋理特征識(shí)別精度最高,可滿足田間雜草識(shí)別要求。
關(guān)鍵詞:玉米;雜草;識(shí)別;紋理特征;灰度矩陣;統(tǒng)計(jì)矩
中圖分類(lèi)號(hào): TP391.41 文獻(xiàn)標(biāo)志碼: A 文章編號(hào):1002-1302(2014)07-0143-03
收稿日期:2013-11-03
基金項(xiàng)目:河北省保定市科技基金(編號(hào):13ZN021);保定學(xué)院科研基金(編號(hào):2013Z04)。
作者簡(jiǎn)介:王懷宇(1975—),男,河北保定人,碩士,講師,從事圖像處理、數(shù)據(jù)挖掘研究。E-mail:why_bdxy@163.com。玉米苗期常見(jiàn)雜草包括刺兒菜、藜、馬唐、田旋花等。傳統(tǒng)除草方法是噴灑除草劑,但田間雜草生長(zhǎng)分布呈不均勻、無(wú)規(guī)律的隨機(jī)分布,因此大規(guī)模藥物噴灑不僅造成浪費(fèi),也對(duì)環(huán)境帶來(lái)不容忽視的污染。隨著精準(zhǔn)農(nóng)業(yè)的發(fā)展和圖像處理技術(shù)應(yīng)用的深入,在機(jī)器視覺(jué)的協(xié)助下實(shí)現(xiàn)農(nóng)田中除草劑的變量噴灑成為當(dāng)今研究熱點(diǎn)。如何識(shí)別雜草圖像是其中最為關(guān)鍵的步驟。當(dāng)前已經(jīng)開(kāi)發(fā)出不少識(shí)別雜草的有效方法,這些方法往往結(jié)合雜草各類(lèi)特征對(duì)其進(jìn)行識(shí)別,包括顏色特征[1-4]、形狀特征[5-6]、光譜特征[7-9]等,也有研究結(jié)合以上組合特征進(jìn)行識(shí)別,取得了較高的效率與精度。但單獨(dú)針對(duì)雜草紋理特征進(jìn)行智能識(shí)別的研究尚不多見(jiàn)。
紋理特征能夠體現(xiàn)出圖像灰度或顏色分布的可描述規(guī)律,尤其是在被識(shí)別目標(biāo)的形狀、顏色等屬性均與周邊環(huán)境相似時(shí),能夠以兼顧宏觀性質(zhì)與細(xì)部結(jié)構(gòu)的方式取得較好的識(shí)別效果[10]。不同作物種類(lèi)或同種作物的健康苗株與病害植株間在圖像紋理特征上有較為明顯的區(qū)別,因此紋理特征在農(nóng)作物病蟲(chóng)草害識(shí)別研究中能取得較好的識(shí)別效果[10]。對(duì)于雜草識(shí)別來(lái)講,怎樣快速提取紋理特征以及如何實(shí)現(xiàn)準(zhǔn)確的識(shí)別率是最關(guān)鍵的問(wèn)題。本研究以玉米常見(jiàn)雜草圖像識(shí)別為例,在圖像預(yù)處理后,對(duì)樣品的多個(gè)紋理特征進(jìn)行篩選,以支持向量機(jī)進(jìn)行分類(lèi),分別進(jìn)行基于灰度矩陣、統(tǒng)計(jì)矩的識(shí)別以及結(jié)合二者的識(shí)別,以期為雜草的快速檢測(cè)及定向施藥提供基礎(chǔ)。
1材料與方法
1.1圖像采集
北方地區(qū)玉米苗期雜草非常常見(jiàn),一般在播種后便可觀察到雜草。據(jù)統(tǒng)計(jì),華北地區(qū)玉米播種后的8~10 d是雜草出土最集中的時(shí)期,12~15 d雜草出土占總量的80%,25 d后達(dá)到95%。雜草的出土、生長(zhǎng)時(shí)間規(guī)律與北方地區(qū)夏玉米苗期生長(zhǎng)節(jié)律基本吻合。因此,只有及時(shí)除去雜草才能保證玉米產(chǎn)量。研究證實(shí),當(dāng)玉米生長(zhǎng)至3~5葉、田間雜草生長(zhǎng)至2~3葉時(shí),是去除雜草的關(guān)鍵時(shí)期。本研究在玉米生長(zhǎng)至3~5葉時(shí)采集田間各類(lèi)雜草圖像,包括刺兒菜、藜、馬唐、田旋花。在目標(biāo)物正上方以640×480像素進(jìn)行拍攝,雜草圖像實(shí)例見(jiàn)圖1。
1.2圖像預(yù)處理
1.2.1圖像增強(qiáng)為了突出圖像特征,削弱某些不重要甚至干擾的信息,首先對(duì)原始圖像進(jìn)行增強(qiáng)處理,以提升圖像中有價(jià)值區(qū)域的對(duì)比度。圖像增強(qiáng)方法分為頻率增強(qiáng)法和空間增強(qiáng)法兩大類(lèi)。考慮到圖像識(shí)別對(duì)于實(shí)時(shí)性的要求,本研究選取效率更高的空間增強(qiáng)法[11],該方法對(duì)每個(gè)像素的灰度值進(jìn)行變換,最終實(shí)現(xiàn)整體對(duì)比度的提升,達(dá)到圖像增強(qiáng)的目的。
3結(jié)論
針對(duì)玉米田間雜草的識(shí)別問(wèn)題,提出了根據(jù)紋理特征進(jìn)行智能識(shí)別的方法,具有創(chuàng)新性。以灰度共生矩陣和統(tǒng)計(jì)矩來(lái)描述識(shí)別目標(biāo)的紋理特征,通過(guò)支持向量機(jī)進(jìn)行分類(lèi)識(shí)別。結(jié)果表明,綜合了灰度矩陣與統(tǒng)計(jì)矩的紋理特征識(shí)別精度超過(guò)90%,能滿足識(shí)別要求。本研究成果有利于減少除草劑噴灑量,有助于實(shí)現(xiàn)除草系統(tǒng)的自動(dòng)化和智能化。
參考文獻(xiàn):
[1]Woebbecke D M,Meyer G E,Von Bargen K,et al. Color indices for weed identification under various soil,residual,and lighting conditions[J]. Transactions of the ASAE,1995,38(1):259-269.
[2]El-Faki M S,Zhang N,Peterson D E. Factors affecting color-based weed detection[J]. Transactions of the ASAE,2000,43(4):1001-1009.
[3]Tang L,Tian L,Steward B L. Color image segmentation with genetic algorithm for in-field weed sensing[J]. Transactions of the ASAE,2000,43(4):1019-1027.
[4]El-Faki M S. Zhang N,Peterson D E. Weed detection using color machine vision[J]. Transactions of the ASAE,2000,43(6):1969-1978.
[5]Guyer D E,Miles G E,Schreiber M M,et al. Machine vision and image processing for plant identification[J]. Transactions of the ASAE,1986,29(6):1500-1507.
[6]Franz E,Gebhardt M R,Unklesbay K B. Shape description of completely visible and partially occluded leaves for identifying plants in digital images[J]. Transactions of the ASAE,1991,34(2):673-681.
[7]Franz E,Gebhardt M R,Unklesbay K B. The use of local spectral properties of leaves as an aid for identifying weed seedlings in digital images[J]. Transactions of the ASAE,1991,34(2):682-687.
[8]Lu R,Peng Y K. Hyperspectral scattering for assessing peach fruit firmness[J]. Biosystems Engineering,2006,93(2):161-171.
[9]Goel P K,Prasher S O,Patel R M,et al. Use of airborne multispectral imagery for weed detection in field crops[J]. Transactions of the ASAE,2002,45(2):443-449.
[10]周平,汪亞明,趙勻. 基于顏色分量運(yùn)算與色域壓縮的雜草實(shí)時(shí)檢測(cè)方法[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2007,38(1):116-119.
[11]何東健. 數(shù)字圖像處理[M]. 西安:西安電子科技大學(xué)出版社,2004.
[12]劉立宏,胡可剛,劉立欣. 目標(biāo)檢測(cè)中的快速中值濾波法[J]. 吉林大學(xué)學(xué)報(bào):信息科學(xué)版,2004,22(3):232-235.
[13]Haralick R M,Shanmugam K. Texture features for image classification[J]. IEEE Systems Man and Cyberbetucs,1973,3(6):610-621.
[14]Ulaby F T,Kouyate F,Brisco B,et al. Textural infornation in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing,1986,GE-24(2):235-245.
[15]鄧乃揚(yáng),田英杰 .數(shù)據(jù)挖掘中的新方法:支持向量機(jī)[M]. 北京:科學(xué)出版社,2004.
[16]Valentini G,Dietterich T G. Bias-variance analysis of support vector machines for the development of SVM-based ensemble methods[J]. Journal of Machine Learning Research,2004,5(7):725-775.
[7]Franz E,Gebhardt M R,Unklesbay K B. The use of local spectral properties of leaves as an aid for identifying weed seedlings in digital images[J]. Transactions of the ASAE,1991,34(2):682-687.
[8]Lu R,Peng Y K. Hyperspectral scattering for assessing peach fruit firmness[J]. Biosystems Engineering,2006,93(2):161-171.
[9]Goel P K,Prasher S O,Patel R M,et al. Use of airborne multispectral imagery for weed detection in field crops[J]. Transactions of the ASAE,2002,45(2):443-449.
[10]周平,汪亞明,趙勻. 基于顏色分量運(yùn)算與色域壓縮的雜草實(shí)時(shí)檢測(cè)方法[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2007,38(1):116-119.
[11]何東健. 數(shù)字圖像處理[M]. 西安:西安電子科技大學(xué)出版社,2004.
[12]劉立宏,胡可剛,劉立欣. 目標(biāo)檢測(cè)中的快速中值濾波法[J]. 吉林大學(xué)學(xué)報(bào):信息科學(xué)版,2004,22(3):232-235.
[13]Haralick R M,Shanmugam K. Texture features for image classification[J]. IEEE Systems Man and Cyberbetucs,1973,3(6):610-621.
[14]Ulaby F T,Kouyate F,Brisco B,et al. Textural infornation in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing,1986,GE-24(2):235-245.
[15]鄧乃揚(yáng),田英杰 .數(shù)據(jù)挖掘中的新方法:支持向量機(jī)[M]. 北京:科學(xué)出版社,2004.
[16]Valentini G,Dietterich T G. Bias-variance analysis of support vector machines for the development of SVM-based ensemble methods[J]. Journal of Machine Learning Research,2004,5(7):725-775.
[7]Franz E,Gebhardt M R,Unklesbay K B. The use of local spectral properties of leaves as an aid for identifying weed seedlings in digital images[J]. Transactions of the ASAE,1991,34(2):682-687.
[8]Lu R,Peng Y K. Hyperspectral scattering for assessing peach fruit firmness[J]. Biosystems Engineering,2006,93(2):161-171.
[9]Goel P K,Prasher S O,Patel R M,et al. Use of airborne multispectral imagery for weed detection in field crops[J]. Transactions of the ASAE,2002,45(2):443-449.
[10]周平,汪亞明,趙勻. 基于顏色分量運(yùn)算與色域壓縮的雜草實(shí)時(shí)檢測(cè)方法[J]. 農(nóng)業(yè)機(jī)械學(xué)報(bào),2007,38(1):116-119.
[11]何東健. 數(shù)字圖像處理[M]. 西安:西安電子科技大學(xué)出版社,2004.
[12]劉立宏,胡可剛,劉立欣. 目標(biāo)檢測(cè)中的快速中值濾波法[J]. 吉林大學(xué)學(xué)報(bào):信息科學(xué)版,2004,22(3):232-235.
[13]Haralick R M,Shanmugam K. Texture features for image classification[J]. IEEE Systems Man and Cyberbetucs,1973,3(6):610-621.
[14]Ulaby F T,Kouyate F,Brisco B,et al. Textural infornation in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing,1986,GE-24(2):235-245.
[15]鄧乃揚(yáng),田英杰 .數(shù)據(jù)挖掘中的新方法:支持向量機(jī)[M]. 北京:科學(xué)出版社,2004.
[16]Valentini G,Dietterich T G. Bias-variance analysis of support vector machines for the development of SVM-based ensemble methods[J]. Journal of Machine Learning Research,2004,5(7):725-775.