基于人工神经网络的玉米籽粒形态识别方法的研究

An Investigation on Morphological Discrimination of Corn Kernels Based on Neural Network

摘要: 该文应用计算机视觉技术选择并获得了11个玉米粒形态参数,采用BP冲量算法,建立了一个三层前馈神经元网络,实现了一幅多颗粒任意放置玉米粒的完整与破损的在线自动识别,对175粒完整及175粒破损玉米粒的识别试验显示,正确率为93％。

Abstract: With 11 selected morphological parameters and BP moment learning algorithm,a three layers feed forward neural network is formed,which can identify more than one whole or broken corn kernels positioned at random in one frame. The classification accuracy of 175 whole and 175 broken corn kernels is 93%.