基于神经网络的图像分割方法及其在颗粒检测中的应用研究

Image Segmentation Based on Neural Network Method and its Application in the Detection of Particles

随着计算机速度的不断提高,数字图像处理技术的发展越来越快。而基于数字图像处理的颗粒检测技术在国计民生中的应用也越来越广,提高颗粒检测技术的速度与准确性显得越来越重要。本文详细介绍了采用模糊竞争Hopfield神经网络对图像灰度级进行聚类,也就是需在某种目标函数最小化的条件下,实现灰度特征集的最优模糊划分,得到最优的阈值,实现图像的分割,并在此基础之上建立了一套对颗粒进行特征参数测量的图像测量系统,可以有效地实现对大量颗粒的测量和表征。

This thesis recommends the particle measurement techniques based on digital image process technology in detail. Set up a characteristic parameter of the particle measurement system on this foundation. The fuzzy competitive hopfield neural networks to cluster the grey grade are adopted in this thesis. Under the target function’s minimum we realize that grey characteristic muster is carved up best. Get the best threshold, and realize image segmentation. On this basis, we established a set of parameters for the measurement of particle image measuring system. It can effectively achieve the measurement and characterization of a large number of particles.