Abstract: In order to evaluate alfalfa (Medicago sativa L.) grade objectively, BP neural network model was established using CP, NDF, ADF, DDM and DMI based on the BP artificial neural network of MATLAB. The neural network model of BP was trained by 200 alfalfa samples. The simulation result indicated that BP neural network model was more suitable to evaluate alfalfa grade and the rate of accuracy was up to 99.6% compared with the artificial result under the characteristic parameter of alfalfa. Therefore, alfalfa quality classification system established in this paper had potential to evaluate alfalfa grade in alfalfa market.

Key words: Alfalfa, Quality, Classification system, MATLAB

摘要： 为了客观评估苜蓿（Medicago sativa L.）草品质的等级，采用MATLAB中BP人工神经网络，利用苜蓿粗蛋白（CP）、中性洗涤纤维（NDF）、酸性洗涤纤维（ADF）、随意可消化量（DDM）、采食量（DMI）参数建立BP神经网络模型。通过200个苜蓿样本进行网络训练，并采用不同的BP神经网络隐含层的传递函数和隐含层神经元数量，获得最优BP神经网络模型。结果表明：在5个特征参数指标下，仿真评价苜蓿草品质等级的准确率达到99.6%，与人工评估结果相比，仿真结果更符合苜蓿草品质的客观现实。在此基础上，介绍已经开发建立的我国首个苜蓿草品质分级系统，有助于未来在苜蓿草市场中发挥其等级评定的应用潜力。

关键词: 苜蓿, 品质, 分级系统, MATLAB