摘要
目标检测是计算视觉的重要研究方向之一,尤其是基于移动设备平台实现快速精准的目标检测功能是非常有必要的。为了能够在移动设备上进行实时目标检测,本文提出一种基于Raspberry Pi 4B硬件平台,采用TensorFlow Lite开发环境,加载MobileNet-SSD网络结构算法的方案。方案采用的MobileNet卷积神经网络和SSD卷积神经网络结合的方法具有检测速度快、占用内存少等优点。同时,本文还对MobileNet-SSD网络结构算法进行了微小优化。该方案通过在公开的数据集上进行测试,对MobileNet-SSD网络结构算法和其改良算法进行了结果比较,结果表明其改良算法的检测速度有所提高,同时其检测精度几乎保持不变,在检测精准度和检测速度上都有良好表现,表明该方案具有较高的应用价值。
Target detection is one of the important research directions of computational vision.Especially,it is necessary to achieve fast and accurate target detection based on mobile device platforms.This paper proposes a solution based on the Raspberry Pi4 B hardware environment which uses the TensorFlow Lite development platform and loads the MobileNet-SSD network structure algorithm.The combination of MobileNet convolutional neural network and SSD convolutional neural network used in the scheme has the advantages of fast detection speed and less memory.At the same time,this paper also optimizes the MobileNet-SSD network structure algorithm.The program compares the results of the MobileNet-SSD network structure algorithm with its improved algorithm through experiments on public data sets,and the results show that the detection speed of the improved algorithm has increased,while the detection accuracy remains almost unchanged.Both accuracy and detection speed have good performance,indicating that the program has high application value.
作者
林向会
谢本亮
LIN Xianghui;XIE Benliang(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)
出处
《智能计算机与应用》
2021年第2期80-83,87,共5页
Intelligent Computer and Applications
基金
贵州省半导体功率器件教育部工程研究中心开放基金项目(ERCMEKFJJ2019-(06))。
