基于图像分割的苹果叶片几何参数计算

针对传统方法和叶面积仪法测定作物叶面积、叶长、叶宽等参数费时费力,成本高等问题,提出苹果叶面积、叶长、叶宽和周长的图像测定法。首先采用颜色空间转换法转换RGB(红、绿、蓝)图为HIS(色调、亮度、饱和度)图,并以大津法(Otsu)阈值色调(H)获取二值图像;再根据二值图中苹果叶边缘像素坐标分布特征,计算叶长、叶宽和叶周长,根据苹果叶片像素数计算叶面积;最后对叶长、叶宽和叶面积的计算值与实测值之间进行均方根误差(RMSE)和决定系数(R~2)计算,并以测定硬币周长的方式验证周长算法。结果表明:苹果叶面积、叶长、叶宽的RMSE值分别为0.58cm~2、0.46和0.10cm,R~2值分别为0.99、0.93、0.97,测定的1元、5角、1角硬币的周长分别为7.68、6.42和6.10cm。RMSE取值较小和R~2取值较高表明叶面积、叶长和叶宽算法结果可靠,周长算法验证结果表明周长计算值与实测值之间差异较小。

Calculating geometrical parameters of apple leaf based on the image segmentation

Given that it is a tedious and high cost work to measure area,length,width and other parameters of crop using traditional methods and leaf area meter,a image method for calculating area,perimeter,length and width of apple leaf was proposed in this paper.Firstly,RGB (red,green,blue) image was convert into HIS (hue,intensity,saturation) color space.Secondly,Otsu method of segmentation was applied to implement binaryzation of H color channel.And finally,the perimeter,length and width of apple leaf were determined based on the distribution feature of boundary pixel coordinate of apple leaf and the apple leaf area was calculated by counting the number of apple leaf pixels.In order to prove the rationality of methods,the RMSE and R² between the calculated and measured apple leaf area,leaf length and width was calculated and perimeter algorithm was test with the help of estimating the circumference of the different size coins.The results showed that RMSE and R² for apple leaf area,leaf length and width were 0.58 cm²,0.46 cm,0.10 cm,0.99,0.93 and 0.97,respectively.And the calculated circumference of one yuan,fifty cents and ten cents of RMB was 7.68 cm,6.42 cm and 6.10 cm,respectively.The low values of RMSE and high values of R² indicated that the result obtained by method of measurement of apple leaf area,leaf length and width were reliable.And the verification results with respect to method of measurement of apple leaf perimeter showed that there were little differences between calculated and true values.