农业网络信息
農業網絡信息
농업망락신식
AGRICULTURE NETWORK INFORMATION
2015年
7期
40-44
,共5页
李宗远%杜金哲%蔡灿超%杨锦忠
李宗遠%杜金哲%蔡燦超%楊錦忠
리종원%두금철%채찬초%양금충
谷子果穗%形态测量%图像处理%不变性%非刚性农产品
穀子果穗%形態測量%圖像處理%不變性%非剛性農產品
곡자과수%형태측량%도상처리%불변성%비강성농산품
millet ear%morphometry%image processing%invariance%non-rigid agricultural product
本文评价了一种能够快速测量谷子果穗形态的数字图像处理技术，以9个品种的谷子作为材料，每品种20个果穗，使用CCD扫描仪采集谷穗的RGB彩色图像。设计算法对RGB图像进行降噪、灰度化、谷穗分割等处理，测量谷子的穗粗，穗长，穗侧面积，和穗体积等4个形态性状，改变谷穗图像的扫描分辨率、亮度、旋转和镜像后，重新测量，测试图像处理算法鲁棒性。同一性状测量值在图像改变前后的直线回归系数分别为，穗体积在0.985~1.075之间，穗侧面积在0.958~1.050之间，穗长在0.9916~1.030之间，穗粗在1.001~1.050之间。全部决定系数值不小于0.992。结果表明新技术具有图像分辨率、亮度、旋转和镜像的不变性，测量精确度高。开发的数字图像处理技术能够快速、精确测量谷子果穗的形态特征，对今后实现对谷子形态测量自动化奠定了基础。
本文評價瞭一種能夠快速測量穀子果穗形態的數字圖像處理技術，以9箇品種的穀子作為材料，每品種20箇果穗，使用CCD掃描儀採集穀穗的RGB綵色圖像。設計算法對RGB圖像進行降譟、灰度化、穀穗分割等處理，測量穀子的穗粗，穗長，穗側麵積，和穗體積等4箇形態性狀，改變穀穗圖像的掃描分辨率、亮度、鏇轉和鏡像後，重新測量，測試圖像處理算法魯棒性。同一性狀測量值在圖像改變前後的直線迴歸繫數分彆為，穗體積在0.985~1.075之間，穗側麵積在0.958~1.050之間，穗長在0.9916~1.030之間，穗粗在1.001~1.050之間。全部決定繫數值不小于0.992。結果錶明新技術具有圖像分辨率、亮度、鏇轉和鏡像的不變性，測量精確度高。開髮的數字圖像處理技術能夠快速、精確測量穀子果穗的形態特徵，對今後實現對穀子形態測量自動化奠定瞭基礎。
본문평개료일충능구쾌속측량곡자과수형태적수자도상처리기술，이9개품충적곡자작위재료，매품충20개과수，사용CCD소묘의채집곡수적RGB채색도상。설계산법대RGB도상진행강조、회도화、곡수분할등처리，측량곡자적수조，수장，수측면적，화수체적등4개형태성상，개변곡수도상적소묘분변솔、량도、선전화경상후，중신측량，측시도상처리산법로봉성。동일성상측량치재도상개변전후적직선회귀계수분별위，수체적재0.985~1.075지간，수측면적재0.958~1.050지간，수장재0.9916~1.030지간，수조재1.001~1.050지간。전부결정계수치불소우0.992。결과표명신기술구유도상분변솔、량도、선전화경상적불변성，측량정학도고。개발적수자도상처리기술능구쾌속、정학측량곡자과수적형태특정，대금후실현대곡자형태측량자동화전정료기출。
This paper aimed at developing and evaluating a digital image processing technology that could quickly measure the ear millet. Taking 9 millet varieties as tested materials, RGB color images of 20 millet ears for each variety were collected using a CCD scanner. The morphological characters of millet, namely ear diameter, ear length, ear side of the area and ear size, were measured by the RGB images, whose noise reduction, graying and ears of corn processing division were optimized by using designed algorithm;What’s more, robustness was remeasured after changing scanning resolution, brightness, rotation and mirroring of ear images. The results showed that the linear regression coefficients between before changed and after changed of the same traits were as follows:0.985-1.075 for side of the ear volume, 0.958-1.050 for ear area, 0.9916-1.030 for ear length, 1.001-1.050 for ear diameter. All decisions coefficient were more than 0.992, which indicted that digital image processing technology had better image resolution, brightness, and rotation and mirroring of invariance, showed higher measurement accuracy. The developed digital image processing technology could quickly and accurately measure the morphological characteristics of millet ear, which laid the foundation for automated measurement of the millet.
