基于混合海鸥算法优化PLSR的果园土壤有效氮含量预测方法

打开文本图片集

Abstract：Nitrogen is one of the three essential elements for crop nutrition and also one of the key nutrient elements under focused monitoring in fertilization management. How to achieve rapid non-destructive monitoring of soil available nitrogen content plays a crucial role in accurate and eficient soil nitrogen application for crops. In this paper，five spectral preprocessing methods were used to remove noise from the spectrum，namely first derivative （Dl），second derivative （D2），standard normal variable （SNV），smoothing gaussian （SG），and wavelet transform （WAVE）. The modeling efects of different spectral preprocessing methods were compared； meanwhile，the competitive adaptive reweighted sampling （CARS） method was adopted to select characteristic bands with high correlation with soil available nitrogen content. Subsequently， prediction models for soil available nitrogen content were established and optimized using support vector regression （SVR）， partial least squares regression （PLSR），and seagull optimization algorithm （SOA），respectively. The results showed that among the five spectral preprocessing methods，the first derivative（D1） preprocessing achieved the best effect. In terms of modeling performance，the SOA algorithm was the most effective，followed by PLSR，while SVR was relatively the worst.Among all models，the S-PLSR model obtained after Dl spectral preprocessing （D1-CARS-S-PLSR model） exhibited the optimal prediction performance，with a coefficient of determination （R2 ）of 0.96 and a root mean square error（RMSE） of 7.O7. Compared with the original PLSR model，the R2 of the D1-CARS-S-PLSR model was increased by 9.6% ，and the RMSE was decreased by 33.3% .It can be concluded that the prediction method for orchard soil available nitrogen content based on PLSR optimized by the hybrid seagull optimization algorithm provides a possibility for the accurate diagnosis of soil available nitrogen levels in the context of orchard intelligent fertilization technology.

Key words： soil available nitrogen content; hyperspectral imaging technology; seagull algorithm； partial least squares regression

柑橘是我国种植面积和产量最大的水果作物，对我国南方农村乡村振兴发挥了重要作用[11。（剩余8911字）