基于计算机视觉技术的玉米叶绿素垂直分布研究

Vertical Distribution of Chlorophyll of Maize Leaf Based on Computer Vision

采用智能手机拍照获取玉米大喇叭口期、抽雄期叶片图像,基于计算机视觉技术计算RGB颜色模型的16种颜色特征参数,探究玉米上、中、下层位叶片颜色特征参数与叶绿素相对含量(SPAD值)的相关关系,并基于随机森林算法建立了SPAD的预测模型。结果表明:两个生育时期玉米叶片叶绿素含量呈垂直异质性,中层叶片叶绿素含量显著大于上层和下层。不同层位叶片叶绿素敏感颜色参数略有不同,但总体上与G、R分量及其组合参数2G-R、2G-R-B、G/R相关性较好,且呈显著负相关。基于RF模型对各层位叶绿素含量进行预测,模型精度表现为上层>中层>下层。基于计算机视觉技术研究玉米叶绿素垂直分布具有可行性,可为及时掌握农作物的光合效率、水肥状况和生长状态,为高产育种和栽培措施调控提供理论依据与技术支撑。

The chlorophyll content of maize leaves and its vertical distribution in the plant well reflect the photosynthetic efficiency,nitrogen stress,water condition and development status of the plant.To reveal the relationships between the color parameters and chlorophyll content(SPAD readings)of the upper,middle and lower maize leaves,digital images of maize leaves at booting and tasseling stages were obtained by smart phones,and 16 color parameters of RGB(Red,Green,Blue)color model were derived by using computer vision technology,and the correlation analysis and the estimation model of SPAD was established by random forest(RF)algorithm.The results showed that chlorophyll contents of maize leaves were vertically heterogeneous at the two growth stages,and the chlorophyll contents of the middle layer leaves were greater than those of the upper and lower layers leaves;the chlorophyll sensitive color parameters at different layers were slightly different,but in general,they presented significant negative correlation with G and R components and their combinations such as 2G-R,2G-R-B and G/R.Based on RF models,chlorophyll contents of each layer leaves were estimated,and the estimation accuracies for different leaf layer by the models were as upper layer>middle layer>lower layer.The results of the research on the vertical distribution of chlorophyll in maize leaves based on computer vision technology provided a theoretical basis and technical support for timely monitoring the photosynthesis efficiency,water and fertilizer condition and growth status of crops,as well as for the high-yield breeding and establishment of cultivation measures.