基于改进人工神经网络植物叶面积测定.PDFVIP

20 13 年2 月 农 业 机 械 学 报 第44 卷 第2 期 doi:10.6041/ j.issn.1000鄄1298.2013.02.037 基于改进人工神经网络的植物叶面积测定* 1 1 1 2 郭孝玉摇 孙玉军摇 王轶夫摇 林静媛 (1.北京林业大学省部共建森林培育与保护教育部重点实验室,北京 100083;2.福建农林大学园林学院,福州350002) 摘要:叶面积作为植物光合作用的重要指标,是研究作物及林木生产力的基础。 采用L M算法和贝叶斯规则相 结合的网络训练模式,以毛竹叶面积为研究对象,综合优化其人工神经网络结构,构建最优的叶面积预测模型。 研 究结果显示,模型的最佳预测变量为叶片宽度和叶片长度变量组合,而增加叶片形状指数未提高叶面积预测模型 精度;所建神经网络模型性能好、预测精度高,决定系数达0郾992,平均相对预测误差为4郾28%,可以准确估测毛竹 叶面积。 关键词:毛竹摇 叶面积摇 人工神经网络摇 贝叶斯规则摇 测定 中图分类号:TP183;S758 文献标识码:A 文章编号:1000鄄1298(2013)02鄄0200鄄05 Improved Artificial Neural Network for Determination of Plant Leaf Area 1 1 1 2 Guo Xiaoyu 摇 Sun Yujun 摇 Wang Yifu 摇 Lin Jingyuan (1.The Key Laboratoryfor Silviculture and Conservation of Ministry of Education,Beijing Forestry University,Beijing 100083,China 2.College of LandscapeArchitecture,FujianAgriculture and Forestry University,Fuzhou350002,China) Abstract: Leaf area is an essential indicator of photosynthesis for the study of crop and forest productivity. The Levenberg Marquardt back鄄propagation optimization algorithm was coupled with Bayesian regulationtotraintheartificialneuralnetwork (ANN),andthepredictivemodelwasdeveloped to determinate rapidly and accurately Moso bamboo leaf area. The results showed that the best input variableswerethecombinationof leafwidth andleaf lengthforANNmodel,whereastheleaf shapeindex did not significantly affect the variability of leaf area. The optimization ANN model possessed with excellent performance and predictable accuracy,with the high determination coefficient of 0郾992 and mean relative prediction error of4郾28%. The ANN model would be allowed for estimating accuracy the leaf area of Moso b