基于Zynq平台的垃圾分类系统实现  

Implementation of waste sorting system based on Zynq platform

作　　者：赵鑫[1] 孟令军[1] 刘威宏 ZHAO Xin;MENG Lingjun;LIU Weihong(Key Laboratory of Instrumentation Science and Dynamic Measurement,Ministry of Education,North University of China,Shanxi Taiyuan 030051,China)

机构地区：[1]中北大学省部共建动态测试技术国家重点实验室,山西太原030051

出　　处：《工业仪表与自动化装置》2022年第5期26-31,96,共7页Industrial Instrumentation & Automation

摘　　要：为了解决GPU上进行硬件加速所导致的功耗较大、成本偏高的问题,该文设计了一套可部署在嵌入式设备中的基于轻量级神经网络MobileNet_v2的垃圾分类系统。该系统将加速的MobileNet_v2网络部署在Zynq7020开发板上,利用HLS工具设计MobileNet_v2网络硬件加速IP核,PS端实现对OV5640摄像头采集并显示图像类别。经过实际测试,该系统可以实现每幅图76 ms的推理速度,能耗比其他平台有了一定的提升,能够较好地满足部署在嵌入式设备的需求。In order to solve the problems of high power consumption and high cost caused by hardware acceleration on GPU,a lightweight neural network MobileNet_v2 based garbage sorting system that can be deployed in embedded devices is designed.The system deploys the accelerated MobileNet_v2 network on a Zynq7020 development board,uses HLS tools to design the MobileNet_v2 network hardware acceleration IP core,and implements the PS side to the OV5640 camera to capture and display image categories.After actual testing,the system can achieve an inference speed of 76 ms per image,and the energy consumption ratio has been improved over other platforms,which can better meet the needs of deployment in embedded devices.

关 键 词：MobileNet_v2 Zynq7020 硬件加速IP核 垃圾分类 HLS 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]