基于L-M神经网络的储粮害虫分类识别

基于图像处理的储粮害虫检测过程中,需要解决多种害虫多特征、混合度大的综合分类问题.本文提出采用基于L-M算法的多层前馈神经网络对害虫进行分类识别.实验表明,该神经网络和害虫分类识别系统拟和程度很高,并且采用的L-M算法,在网络训练速度及识别精度方面,都优于传统的BP算法.因此基于L-M算法的神经网络在害虫的在线识别方面有应用价值.

Stored-grain pests classification based on L-M neural networks

Using image processing technology to recognize the stored-grain pests,the classification and recognition based on multi-characteristic parameters and multi-compound degree of various pests must be settled.In this paper,an artificial neural network based on Levenberg-Marquardt training algorithm is proposed to solve this problem.According to the result,the neural network can simulate the recognition system very well,and furthermore,the Levenberg-Marquardt algorithm is superior to the classic back propagation(BP)algorithm ether from the training speed or from the accuracy.Therefore,a neural network based on L-M algorithm can be applied to the pests' Real-time recognition system.