基于流水线图像处理结构的多层模板相关神经元网络

在自主地面车辆中，视觉系统的重要作用之一是根据路标来定位。本文提出了一种便于流水线图像处理结构实现的多层模板相关神经元网络（ＭＴＣＮＮ）。文中给出了ＭＴＣＮＮ的基本结构及训练算法，并且将其与经典的多层前馈神经元网络（ＭＬＦＮＮ）进行了比较。仿真结果表明，本文提出的算法结构在多层前馈神经元网络的分类能力与采用通用图像处理硬件的可实现性之间，取得了良好的折衷。

Multi-layer Template Correlation Neural Network Based on Pipelined Image Processing Structure

It is one of the important tasks of the vision system of an autonomous land vehicle (ALV) to locate itself by lane mark. In this paper, a multi layer template correlation neural network (MTCNN)based on the pipelined image processing structure is proposed for the recognition of lane mark. A structure of the MTCNN and the training algorithm are presented. In addition, a comparison between MTCNN and MLFNN is introduced. The results of simulation manifest that the proposed MTCNN is very efficient for the task such as recognition of lane mark that is based on the pipelined image processing structure.