基于Hopfield神经网络的图象处理

Image Disposition of Based Hopfield’s Neural Networks

本文应用Ｈｏｐｆｉｅｌｄ网络模型解决计算机视觉中图象匹配与识别问题．对于单幅图象物体的３条边与模型中３条对应边，通过求解一元四次方程求出三维物体姿态．进一步应用Ｈｏｐｆｉｅｌｄ网络求出最佳匹配点，从而识别三维物体．本论文的方法将为计算机视觉中高层次匹配问题提供一种新的并行处理的途径．

The paper comprehensively analysizes the primary model of the neural networks-the Hopfield Networks, and uses it to solve the image matching problem in computer vision. The energy function of Hopfield Networks can decrease constantly in moving network when it satisfies specified parameter conditions, finaly it can reach a stable balance. We can use this energy function as a computing tool for the network obtaining the pose of a 3D object by solving a forth degree equation with one unknown quantity. By matching the three lines of the single image and three lines of the model,and by using Hopfield Network, we obtain the best matching point, and recognize the 3D object. Thus the paper provides a new synchronized process for advanced matching problems in Computer Vision.