基于无人机遥感的花生叶片叶绿素含量监测研究

Research on Monitoring of Peanut Leaf Chlorophyll Content Based on UAV Remote Sensing

为实现对花生叶片叶绿素含量进行快速、实时、无损、准确监测,本研究利用大疆精灵4号无人机搭载可见光相机,获取花生不同生育期的遥感影像,利用深度神经网络(CNN)和卷积神经网络(DNN),建立叶片数字图像彩色信息和叶片叶绿素含量的关系模型。结果表明,利用数字图像特征(包括红光,绿光,蓝光,归一化红光,归一化绿光等)作为网络输入向量时效果较好,CNN产生的预测值和农学实测值训练集平均绝对误差为0.8左右,DNN产生的预测值和农学实测值训练集平均绝对误差为1左右,两者均在误差范围内,可用来实时监测花生的叶绿素含量,为生产上快速诊断花生的肥料状况提供理论依据。

In order to achieve fast,real-time,non-destructive,and accurate monitoring of the chlorophyll content of peanut leaves,this research obtained remote sensing images of peanut at different growth stages by the DJI Genie 4 drone which equipped with a visible light camera.The relation model between digital color image and the chlorophyll content were established through the deep neural network(CNN)and convolutional neural network(DNN).The results showed that the use of digital image features(including red light,green light,blue light,normalized red light,normalized green light,etc.)as the network input vector had the good effect.The average absolute error between the prediction value generated by CNN and the agronomic observed value training set was about 0.8.The average absolute error between the predicted value generated by DNN and the agronomic observed value training set was about 1.The above two monitoring model were both within the error range,and could be used to monitor the chlorophyll content of peanut leaves in real time.This results will provide the theoretical basis for the rapid diagnosis of fertilizer status in peanut production.