基于模糊神经网络的大米品种识别算法研究

摘要

Since grains are probably connected with each other in practical rice images, this paper designed fuzzy neural network to identify the variety of rice based on machine vision. Morphological algorithms including corrosion and expansion are used to extract the single grain. After removing the broken grains, we can obtain the surface characteristics such as long axis,short axis, long axis/short axis, equivalent diameter, and area of rice. The fuzzy BP neural network is applied to do identification. Theory analysis and simulation result show that the method can achieve better detecting precision.%基于机器视觉技术,针对实际大米图像中不可避免的存在籽粒连接的情况,采用模糊BP神经网络进行大米品种识别.应用形态学中腐蚀、膨胀算法提取单粒大米籽粒,去除碎米后,对单粒整精大米籽粒进行外观特征提取,采用模糊BP神经网络进行大米品种识别,仿真结果表明,其可达到较高的检测精度.