基于忆阻神经网络的图像边缘检测的FPGA硬件实现

FPGA Hardware Implementation of Memristor Pulse Coupled Neural Network for Edge Detection Images

摘要: 基于忆阻器的高集成度和非易失性等特点，本文将忆阻器应用于传统神经网络模型中，对原有的脉冲耦合神经网络进行创新，将两个忆阻元件反向并联替代神经网络连接权值，形成忆阻神经网络。同时利用忆阻神经网络可对图片实现边缘检测的特性，将边缘检测结果图与FPGA结合，通过串口通信，SDRAM存储，TFT显示等模块，使用Verilog语言对各个硬件模块进行描述，最后成功在TFT显示屏上显示出结果图。

Abstract: Based on the high integration and non-volatility of memristors, this paper applies memristors to the traditional neural network model, innovates the original pulse-coupled neural network, and re-places the neural network connection weights of two memristive elements in reverse parallel to form a memristive neural network. At the same time, the memristive neural network can be used to realize the edge detection characteristics of the picture, the edge detection result graph is combined with the FPGA, through serial port communication, SDRAM storage, TFT display and other modules, each hardware module is described in Verilog language, and finally the result graph is successfully displayed on the TFT display.