基于PROSAIL模型和遗传算法优化的BP神经网络模型的不同大豆种群叶面积指数反演比较

Inversion comparison of leaf area index of different soybean populations based on PROSAIL model and BP neural network optimized by genetic algorithm

以大豆叶面积指数(Leaf area index,LAI)反演为研究目标,利用PROSAIL模型和遗传算法优化后的BP神经网络模型,分别对重组自交系(Recombinant Inbred Lines,RIL)和自然野生大豆种群的LAI进行反演。结果表明,在对人工定向培育的RIL大豆种群冠层叶片LAI反演中,PROSAIL模型表现出了更优异的反演能力,而对品种繁多的自然野生大豆种群LAI反演中,遗传算法优化后的BP神经网络模型表现出了更好的适用性,并且上述2种模型在始粒期(R5)时性能最佳,PROSAIL模型和遗传算法优化后的BP神经网络模型R2分别为0.89和0.85,RMSE分别为0.11和0.13,EA均为97%,典型生育期内的反演性能均优于全生育期综合反演性能。因此,针对同一农作物不同种群的表型特征反演,需要根据研究对象的特征来选择合适的模型,以便于精确的估测大豆长势情况,为农作物的规模化育种监测提供数据支持。

In this paper,we take the soybean leaf area index(LAI)inversion as the research objective,the PROSAIL model and BP neural network optimized by genetic algorithm were used to invert LAI of Recombinant Inbred Lines(RIL)and natural wild soybean populations.The results demonstrated that PROSAIL model showed excellent inversion ability for LAI of soybean canopy leaves in RIL population,and the optimized BP neural network after genetic calculation had better applicability to natural wild soybean population,and the above two models have the best performance at the initial granular period(R5),with the certainty coefficients of 0.89 and 0.85 for the PROSAIL model and BP neural network optimized by genetic algorithm model.The root mean square errors are 0.11 and 0.13 respectively,and the estimation accuracy of the two models are both 97%.The inversion performance during the typical growth period is better than the comprehensive inversion performance during the entire growth period.Therefore,for the inversion of phenotypic characteristics of different populations of the same crop,it is necessary to select the appropriate model according to the characteristics of the research object,in order to accurately estimate the growth situation of soybean and provide data support for large-scale breeding monitoring of crops.