Planting Area Extraction And Growth Monitoring Of Sugarcane Based On NDVI Time Series

Sugarcane is one of the main source crops of sugar.Timely and accurate acquisition of sugarcane planting area and growth status can help the government and relevant enterprises to formulate sugarcane production policies and assist relevant departments to formulate sugar export plans.The traditional method of obtaining sugarcane planting area is not only timesensitive,but also can not accurately obtain sugarcane spatial distribution.In recent years,great progress has been made in the extraction of sugarcane planting area by remote sensing technology,but there are still some urgent problems to be solved,such as the phenomenon of "foreign bodies with the same spectrum" in the extracted sugarcane pixels.In this study,based on the spectral similarity principle and threshold decision tree method,the classification error of "foreign bodies in the same spectrum" was eliminated to achieve high-precision extraction of sugarcane planting area.The results of sugarcane extraction were used to monitor sugarcane growth and provide data support for predicting sugarcane yield.This research has important theoretical significance and practical value.The main research contents and results of this paper are as follows:(1)Based on the spectral similarity principle and threshold DT method,this paper constructed a high-precision extraction model of sugarcane planting information.Normalized Difference Vegetation Index(NDVI)time series curves were decomposed and reconstructed using singular value decomposition and spectral reconstruction(SR-SVD).Spectral Angle Mapping(SAM),determination coefficient R and Euclidean Distance(ED)were used to evaluate the similarity of NDVI time series before and after pixel reconstruction to extract sugarcane planting information.The experimental results show that using SAM size to evaluate the similarity of NDVI time series before and after pixel reconstruction can effectively capture the changing trend between curves before and after pixel reconstruction,thus highlighting the difference between sugarcane pixel and non-sugarcane pixel.The overall accuracy of extracting sugarcane planting information by SR-SVD-SAM was 88%,which verified the effectiveness of extracting sugarcane by spectral similarity principle.(2)In order to further improve the extraction accuracy of sugarcane planting area,this study used prior knowledge to determine the threshold range of sugarcane NDVI time series curve in different phases and set a threshold decision tree(DT).The experimental results showed that the threshold decision tree(DT)could effectively eliminate the misclassified pixels of "same spectral foreign objects" in the sugarcane extraction results of SR-SVD-SAM,and significantly improve the accuracy of sugarcane extraction.The overall accuracy of extracting sugarcane planting information using SR-SVD-SAM-DT model reached 96%,which has a great advantage over that of extracting sugarcane planting information only using SR-SVDSAM.This model can extract sugarcane planting information quickly and accurately.(3)The growth characteristics of perennial sugarcane and newly planted sugarcane were analyzed,and the change rule of NDVI time series was compared to obtain the best time phase for monitoring sugarcane growth,and the growth trend was monitored and evaluated.Finally,the accuracy of evaluation results was verified by the yield per unit area information of sugarcane yield sample points.The results showed that the growth characteristics of persistent sugarcane and newly planted sugarcane were different in different growth periods.Compared with the results of monitoring sugarcane growth using multi-temporal NDVI,the results of monitoring sugarcane growth using single-temporal NDVI were more consistent with the growth law and logic of sugarcane.Conclusion It provides a convenient and efficient technical means for relevant departments to obtain sugarcane growth information.