杨占魁1,2，任 东1，孙传恒2，周 超2，解 菁2
（1 三峡大学计算机与信息学院，湖北 宜昌 443002；
2 国家农业信息化工程技术研究中心，农业部农业信息技术重点实验室，北京 100097）

摘要：针对人工估算半滑舌鳎鱼苗体重存在误差大、难度大、易伤鱼苗等缺点，提出了一种基于遗传BP神经网络体重估算模型优化研究的方法。首先利用参考系测量鱼苗体长和体宽，再通过遗传BP神经网络估算模型计算鱼苗体重。利用遗传算法对编码后的BP神经网络进行优化并验证了遗传算法能有效确定BP神经网络隐藏层节点数目。结合遗传算法优化BP神经网络的结构和连接权值，采用300份同一训练样本对优化的BP神经网络进行训练，最终建立准确遗传BP神经网络体重估算模型。结果显示，该方法对鱼苗体重估算与实际值平均相对误差不超过0.61%。研究表明，该方法为半滑舌鳎体重估算提供了一种比较科学的计算方法，在鱼苗生长发育监测和科学喂养等方面具有重要的实际意义。

关键词：半滑舌鳎；平均体重估算；遗传算法；BP神经网络

Research on optimization of Cynoglossus semilaevis weight estimation model based on genetic algorithm and BP neural network
YANG Zhankui1,2, REN Dong1, SUN Chuanheng2, ZHOU Chao2, XIE Jin2
(1. Department of Computer and Information Technology, Three Gorges University, Yichang 443002, China;
2. Key Laboratory of Information Technology in Agriculture, Ministry of Agriculture, National Engineering Research Center for Information Technology in Agriculture, Beijing 100097, China )

Abstract：As there are some disadvantages of large error, difficult to estimate, easily hurt the fry for the artificial estimation of Cynoglossus semilaevis fry weight, a optimization method for weight estimation model based on genetic BP neural network was presented in this paper. Firstly the length and width of fish were measured by the reference and then the fish weight was estimated by Genetic BP neural network estimation model. Using genetic algorithm to optimize the structure and weight of encoded BP neural network, the genetic algorithm was proved to be effective in determining the hidden layer nodes of BP neural network. The structure and connection weights of neural network were optimized by genetic algorithm, then the neural network was trained by the same 300 training samples, and finally the accurate BP neural network weight estimation model was established. The results showed that the average relative error was less than 0.61% between the estimated value and the real value. The research suggests that this method provids a scientific calculation method for estimation of Cynoglossu semilaevis fry weight and it plays an important significance in the fry growth monitoring and scientific feeding.

Key words: Cynoglossus semilaevis; average weight estimation; genetic algorithm; BP neural network