采用灰板校正的计算机视觉预测棉花叶绿素含量

为了提高计算机视觉技术对棉花叶绿素含量的预测精度，该文应用计算机视觉识别方法，采用灰板校正图像亮度差异，对不同水分背景下棉花叶片叶绿素含量进行预测。结果表明，光谱特征参数DGCI (dark green color index)、R-B与叶绿素含量之间存在极显著线性关系，未使用灰板校正图像的DGCI、R-B与叶绿素含量的相关系数分别为0.8857和-0.8726，使用灰板校正归一化处理后的相关系数分别为0.9073和-0.9016，灰板校正后提高了参数与叶绿素含量的相关性。比较参数DGCI、R-B在校正前后对叶绿素含量的预测精度，结果显示校正后的DGCI、R-B建立的模型预测精度高于校正前，校正后参数DGCI的预测精度大于R-B。采用校正后参数DGCI建立的Chl. a+b预测方程，其预测值与叶绿素实测值间均方根误差和相对误差分别为0.1200和5.28%，决定系数为0.8812，预测精度较高。应用计算机视觉技术预测不同水分处理下棉花叶绿素含量具有可行性，使用灰板校正后参数DGCI可以作为快速无损预测棉花叶绿素含量的最佳参数。

Estimation of chlorophyll contents in cotton leaves using computer vision based on gray board

Abstract: This paper was an attempt to develop a low-cost quick method that is easy to use to assess the chlorophyll content of cotton plants using color characteristic parameters which was adjusted by a grey board from plant images. The cotton plant images were obtained at different growth periods from different water treatments. Images were adjusted and normalized by a grey board. The second leaf from the top of a cotton stem image was taken by a CMOS digital camera. The camera lens maintained a 90 degree angle with the cotton leaf vertical to shoot. Before taking an image, the location of the camera lens and the cotton leaves was fixed, and the focal length of the lens was fixed. The gray board image was taken first every time. Using the RGB and HSB color system to split the cotton leaf color characteristics, red value (R), green values (G), blue value (B), and Hue (H), saturation (S) and brightness (Br) of the cotton image were obtained through cotton leaf analysis software that was developed by the VB. Chlorophyll content of the cotton leaf was obtained by spectrophotometer determination. The correlation analysis was set up between color characteristics parameters and chlorophyll content. The correlation coefficients between DGCI (dark green color index) or Red-Blue which were not adjusted by grey broad and cotton chlorophyll were 0.8857 or -0.8726, and they were 0.9073 or -0.9016 respectively after correction. The correlation coefficient between parameters and chlorophyll content were improved after grey board adjustment. The result showed that there were a series of color parameters combination obtained from cotton leaf images that were a highly significant linear correlation with chlorophyll content of the cotton leaf in various growth periods. The color characteristic parameters and chlorophyll content in different periods were combined, and the correlation between them was analyzed. DGCI and Red-Blue had the most highly significant linear correlation with cotton leaf chlorophyll content. Comparing the chlorophyll content prediction accuracy of DGCI or Red-Blue before and after the correction, it showed that the parameter DGCI or Red-Blue after adjustment model prediction accuracy is higher than before calibration. The prediction accuracy of DGCI is higher than Red-Blue parameters after calibration. The prediction for DGCI after adjusted was Chl.a+b= 8.3265DGCI-2.0456. Between the predicted values, which were calculated by the equation, and the measured values of chlorophyll, its root mean square errors (RMSE) was 0.1200, and the relative errors (RE %) was 4.71%. The decision coefficient was 0.8812. The prediction accuracy was better. Our results demonstrated that the adjusted DGCI was the best indicator to predict cotton leaf chlorophyll content, and the prediction model was feasible for applying computer vision technology to rapidly predict cotton chlorophyll content.