基于参数优化支持向量机的水稻施氮水平分类研究

目的]应用参数优化支持向量机对水稻施氮水平进行准确分类预测,为水稻精准施肥和高产管理提供科学依据.方法]以水稻品种金优458为试验材料,设4个施氮水平(从高至低折合纯氮用量分别为225、150、75和0 kg/ha),通过叶绿素测量仪SPAD-502获取水稻第6~9叶序叶片的SPAD值(即叶尖、叶中和叶枕的SPAD值),并分别应用网格搜索算法和粒子群算法参数优化支持向量机对4个施氮水平下的水稻叶片SPAD值进行训练和预测分类.结果]对于第7、8叶序、第7~9叶序及第6~8叶序叶片组合,粒子群算法参数优化支持向量机对水稻施氮水平的分类识别效果均优于网格搜索算法,其准确率均高于75.000%,对归一化处理后的第7、8叶序叶片组合识别率最高,达88.889%.结论]基于粒子群算法参数优化支持向量机适用于水稻施氮水平分类预测,能满足农学研究的需求.

Classification of nitrogen application level for rice based on support vector machine optimized by parameters

Objective]Support vector machine optimized by parameters was applied to predict classification of nitro-gen application level for rice in order to provide scientific basis for accurate fertilization and high yield management of rice.Method]Four nitrogen application levels(from high to low,the amount of pure nitrogen was 225,150,75 and 0 kg/ha respectively)were set,and rice cultivar Jinyou 458 was used as experiment material. The SPAD values of the 6th to 9th phyllotaxis rice leaves were obtained by chlorophyll meter SPAD-502(SPAD value of leaf top,leaf middle and leaf bot-tom). The SPAD values of rice leaves under four nitrogen application levels were trained and predicted by using support vector machine optimized by particle swarm optimization and grid search algorithm.Result]For the 7th and 8th phyllotaxis leaf combination,the 7th,8th and 9th phyllotaxis leaf combination and the 6th,7th and 8th phyllotaxis leaf combination,the rice nitrogen application rate classification detected by support vector machine optimized by particle swarm optimization was better than support vector machine optimized by grid search algorithm,its accuracy was 75.000%higher. Moreover, its accuracy on the 7th and 8th phyllotaxis normalized leaf combination was the highest(88.889%).Conclusion]Support vector machine optimized by particle swarm optimization is suitable for predict the classification of rice nitrogen applica-tion levels and meets the needs of agricultural research.