UAV RGB Image-based Maize Growth Monitoring

In recent years,China is in urgent need of real-time accurate monitoring of field crop growth in the context of promoting smart agriculture and precision agriculture.Most of the current researches on crop growth monitoring are based on one kind of growth indicators for crop growth monitoring,thus ignoring other growth indicators of crops.In the practical application of plant growth monitoring,most of them use multispectral and hyperspectral,but their high cost and complicated post-processing make the UAV remote sensing monitoring technology cannot be better applied.To this end,this thesis takes spring maize in Hangjinhou Banner of Inner Mongolia as the research object,uses UAV RGB images and RTK technology to extract maize plant height to establish maize plant height monitoring model,and analyzes maize yield differences by calculating maize Zhu plant height neatness.The soil background in the remote sensing images was removed by supervised classification method,and the spectral index and texture features of each test plot were obtained.The correlation between the spectral index and texture features and the growth indexes(LAI,SPAD,above-ground biomass)were analyzed separately,and the spectral index and texture features with better correlation with the growth indexes were selected as the variables for establishing the model,and the filtered spectral index,texture features and the two We also compared the accuracy of different modeling methods and the results of using spectral indices or texture features alone and using a combination of them to predict the growth indexes;finally,we obtained the best estimation method.The best estimation method was obtained.The main results of the study are as follows.1.The results showed that the accuracy of maize plant height monitoring model in the test site was good,and the R2 of the estimated and measured maize plant height values were above 0.78,and the RMSE was less than 0.21 m,which showed high accuracy.By calculating the plant height neatness among different replications in each plot,the plant height neatness among the same treatments was positively correlated with the final yield of maize,indicating that plant height neatness is also an important indicator of yield.2.The growth indices(LAI,SPAD,aboveground biomass)were in good agreement with the calculated spectral indices and extracted texture features,and all spectral indices except g and EXG were highly significantly correlated with each growth index.The growth indices were positively correlated with the texture features hom(homogeneity),sm(angular second order moment),and cor(correlation),and negatively correlated with mean,var(variance),con(contrast),dis(heterogeneity),and ent(entropy),and all 24 texture features were highly significantly correlated with each of the growth indices.3.The results of maize LAI,SPAD,and aboveground biomass monitoring models using PLSR,SVM,and RF based on spectral indices,texture features,and spectral indices fused with texture features,respectively,were seen.The prediction results of the models using spectral indices fused with texture features were more accurate than those using spectral indices or texture features alone,and the prediction results were better than those of the single spectral index or texture feature models for all three growth indicators.4.Comparing the three different modeling approaches,it was found that the best prediction was achieved for the RF model of maize LAI,SPAD,and aboveground biomass,and the highest accuracy was achieved when the monitoring model was built using spectral indices fused with texture features,with the validation set R2 ranging from0.79 to 0.85 for the three indicators;RMSE ranging from 0.33 to 2.21;and RPD ranging from 2.17 to 2.38.Among them,the RF model can monitor the three growth indicators of maize LAI,SPAD,and aboveground biomass very well.Based on the above results,the UAV-based RGB image technology can realize the monitoring of maize growth,and the fusion of texture features and spectral features can effectively improve the accuracy of the monitoring model.The difference in maize growth is an important factor causing the difference in final yield,and it is important to improve the accuracy of maize growth monitoring to guarantee the maize yield.