草业学报2012，Vol.21Issue(4)：275-281,7.
基于BP和RBF神经网络模型的草坪质量综合评价
A comprehensive evaluation of turfgrass quality based on a BP and RBF neural network model
摘要
Abstract
Based on the recent turfgrass quality evaluation system, eleven indexes (density, texture, color, uniformity, green period, disease resistance, coverage, traffic tolerance, seedling establishment, turf strength and biomass) were used to select 20 Poa pratensis cultivars in 2010. The values of eleven indexes from 15 of the P. pratensis cultivars were selected as input data for the system using the principles of neural networks and the Matlab neural network toolbox. The output was expert graded data. Performance optimization was carried out by running the neural networks with different parameters and then models of the BP and RBF neural networks for evaluation of turfgrass quality were established. Methods of establishing neural network models and steps for Matlab are listed. Quality of the other 5 P. pratensis cultivars was evaluated using the trained neural network model. The predicted errors of the RBF neural network were less than 2% and the predicted errors of the BP neural network were more than 5 % when they were applied as a comprehensive evaluation of turfgrass quality. Therefore the RBF neural network with smaller error odds was able to provide a more accurate evaluation of turfgrass quality than the BP neural network and it can be used to evaluate turfgrass quality. Compared with traditional methods, such as the weighting method, analytic hierarchy process, and fuzzy synthesis, the RBF neural networks accuracy reduces the influences of subjective factors and simplifies the calculating procedures. It provides a new idea for comprehensive evaluation of turfgrass quality.关键词
草坪质量综合评价/BP神经网络/RBF神经网络Key words
comprehensive evaluation of turfgrass quality/BP neural network/RBF neural network分类
农业科技引用本文复制引用
肖波,宋桂龙,韩烈保,包永霞,李飞飞,陈爱霞..基于BP和RBF神经网络模型的草坪质量综合评价[J].草业学报,2012,21(4):275-281,7.基金项目
国家林业局"948"项目(2011-4-50)和北京市重点学科建设项目资助. （2011-4-50）
