Inversion Of Leaf Area Index Of Northern Rice Based On Machine Learning

Rice is one of the three major food crops in the world.Its output and quality play a very important role in the world's food security.The leaf area index is important indicator for rice growth assessment.How to quickly and accurately obtain different physiological parameters of rice is of great significance for scientific fertilization and efficient field management.The purpose of this study is to take the northern rice leaf area index as the research object,two paddy field with different nitrogen levels are chosen as the experiment fields.Based on the ground actual measurement data and the multispectral remote sensing image data,the correlation between the spectral reflectance and the leaf area index are constructed with the traditional statistical analysis and the machine learning regression methods,that can be used to monitor rice growth in the area.The research results can provide more reliable data and technical support for scientific fertilization in precision agriculture.The paper is closely related to the inversion of rice leaf area index.The main research contents and innovations are as follows:(1)Firstly,the growth agronomic basis of rice and the specificity of rice leaf area index are systematically analyzed,and the spectral reflectance characteristics of the rice is also discussed.The results show that the spectrum curve of rice has a clear trend,and the most relevant bands are the green band,the red band,the red edge band and the near-infrared band,Six bands are chosen to construct the invert for rice leaf area index.(2)Secondly,two vegetation indexes,including the ratio vegetation index and the normalized difference vegetation index,are calculated with the spectral reflectance obtained from the Sentinel-2 remote sensing images.The extracted six bands of pixel values and the measured rice leaf area index are used to construct the training data set and the test data set.Then the traditional empirical formula method,including the linear regression,the logarithmic regression and the exponential regression methods,are used to build the regression relationships between the spectral reflectance and the measured leaf area index.The inversion accuracies are also discussed.(3)Finally,two machine learning algorithms,support vector machine and random forest are used to construct the inversion model with the training data set,and the inversion accuracies are tested with the testing data set.The final results show that the multi-band inversion model based on the support vector machine algorithm works best.