首页 > 分享 > 基于时序PolSAR影像与决策树模型的油菜物候期识别

基于时序PolSAR影像与决策树模型的油菜物候期识别

花匠小妙招
2025-06-18 13:54

摘要：

作物物候期识别是农情遥感监测的重要内容,及时准确识别作物物候期,对有效评估作物生长趋势、提高农情信息化管理水平有重要意义。提出了基于时间序列全极化合成孔径雷达(polarimetric synthetic aperture radar, PolSAR)数据结合决策树模型的油菜物候期识别方法。首先,采用3种极化分解方法提取PolSAR极化参数,并分析各极化参数对油菜物候期的动态响应规律;其次,基于各极化分解方法提取的参数建立决策树模型,并对油菜物候期进行分类识别;最后,采用基于混淆矩阵的方法对油菜物候期识别结果进行精度评价。采用5期Radarsat-2 PolSAR数据和地面物候观测数据进行实验验证。结果表明:提取的PolSAR参数中对物候期变化较为敏感的参数有H/A/alpha分解中的散射角(Alpha)、特征值(L2、L3)、伪熵(P2)、目标方位角(Beta1)参数,Freeman-Durden分解中的地面散射(Ground)和奇次散射(Odd)参数,Yamaguchi分解中的奇次散射(Odd_Y)和螺旋体散射(Helix)参数;决策树模型对油菜物候期识别结果较为准确,识别结果中组合3种极化分解方法提取参数建立的原始决策树模型分类总体精度最高,达94%。总体上,PolSAR极化分解参数对油菜物候期变化比较敏感,决策树模型能有效识别油菜物候期。

关键词: 油菜, 物候期识别, Radarsat-2, 极化分解, 决策树

Abstract:

Crop phenological period identification plays an important role in agricultural condition monitoring. Timely and accurate crop phenological period identification is of great significance for evaluating crop growth trend and improving the agricultural information management level effectively.In this study, we selected oilseed rape as an example and proposed a crop phenological period identification method through polarimetric synthetic aperture radar (PolSAR) data and decision tree algorithms. First, polarimetric SAR parameters were extracted through three popular polarimetric decomposition methods. Their dynamic responses to oilseed rape phenological periods were also analyzed. Then, the parameters extracted by the three polarimetric decomposition methods were used to train and validate the decision tree models, five oilseed rape phenological periods were identified. Finally, confusion matrices were used to verify feasibility of the constructed decision models.The results showed that polarimetric SAR decomposition parameters, including scattering angle(Alpha), eigenvalue(L2, L3), pseudo-entropy(P2) and target azimuth (Beta1) parameters from H/A/alpha decomposition method; ground scattering (Ground) and odd scattering (Odd) parameters from Freeman-Durden decomposition method; odd scattering (Odd_Y) and helix scattering (Helix) parameters from Yamaguchi decomposition method showed great sensitivity to changes of oilseed rape phenological period.The decision tree models were more accurate to classify the phenology of rapeseed. Among the results, primitive decision tree model established based on the combination of extracted parameters from three polarimetric decomposition methods had the highest classification accuracy, the overall classification accuracy was 94%. The results also showed the sensitivity of PolSAR parameters to phenological changes of oilseed rape and the effectiveness of decision tree model in identification of oilseed rape phenological period.

Key words: rape, phenological identification, Radarsat-2, polarimetric decomposition, decision-tree

中图分类号:

S565.4

TP75