基于时序Sentinel-2数据的马铃薯遥感识别研究

Research on potato remote sensing recognition based on time series Sentinel-2 data

关键词:  马铃薯  波段  生育期  遥感识别

Key words:  potato  band  growth periods  remote sensing identification

摘要:

[目的] 文章以黑龙江省北大荒农垦集团克山农场为研究区域，以当地马铃薯全生育期为研究时段，以早熟马铃薯、晚熟马铃薯及其他主要农作物大豆、玉米、水稻为研究对象，开展基于时序Sentinel-2数据的识别差异性研究。[方法] 基于早熟马铃薯、晚熟马铃薯及大豆、玉米、水稻各时期各波段光谱反射值，利用光谱差异性指数，创建了光谱振幅指数，结合各种作物生育期数据，评价不同波段、不同时间序列对早熟马铃薯、晚熟马铃薯识别的差异性；筛选出可以较好区分早熟马铃薯、晚熟马铃薯、大豆、玉米、水稻的波段和时间序列组合。[结果] 在该文的时序内，能够同时区分早熟马铃薯、晚熟马铃薯与大豆、玉米、水稻的Sentinel-2数据时段为2022年8月9日和8月31日，波段为Band 5、Band 7和Band 11。利用计算机自动分类，结合人工目视解译方式获得2022年克山农场种植结构，马铃薯、大豆、玉米、水稻、其他作物比例为13∶57∶27∶0∶2（其中的水稻比例不足0.30%）。[结论] 综合Sentinel-2数据各波段反射值、光谱差异性指数，结合生育期资料，使用单一时相识别马铃薯时，最佳时段为晚熟马铃薯的成熟期后，地面光谱表现为裸地，其他作物光谱仍然表现为作物。

Abstract:

[Purpose] This paper takes Keshan Farm of Beidahuang Agricultural Reclamation Group in Heilongjiang Province as the research area，takes the entire growth period of potatoes as the research period，and takes early maturing potatoes，late maturing potatoes，and other major crops such as soybeans，corn，and rice as the research objects to research the identification differences based on temporal Sentinel-2 data.[Methods] Based on the spectral reflectance values of early maturing potatoes，late maturing potatoes，soybean，corn，and rice at different stages，a spectral amplitude index was created using spectral difference indices. Combined with data from various crop growth stages，the differences in recognition of early maturing potatoes and late maturing potatoes in different bands and time series were evaluated；Then，screen out the band and time series combinations that can distinguish early maturing potatoes，late maturing potatoes，soybeans，corn，and rice better.[Results] During the time series of this study，Sentinel-2 data on August 9 and August 31，2022 can simultaneously distinguish early maturing potatoes and late maturing potatoes from soybeans，corn，and rice，with bands of Band 5，Band 7，and Band 11. The planting structure of Keshan Farm in 2022 was obtained by using computer automatic classification combined with manual visual interpretation. The proportion of potatoes，soybeans，corn，rice，and other crops was 13∶57∶27∶0∶2（with rice accounting for less than 0.30%）.[Conclusions] When using a single temporal recognition method to identify potatoes based on Sentinel-2 data，including reflectance values，spectral diversity indices，growth period data，the optimal time period is after the mature stage of late maturing potatoes. The ground spectrum shows bare land，while other crop spectra still show crops.