• 首页
  • 期刊简介
  • 编委会
  • 投稿指南
  • 收录情况
  • 杂志订阅
  • 联系我们
引用本文:黄 静,张 健.一种基于改进SSD网络的猪个体目标检测方法研究[J].软件工程,2022,25(8):25-29.【点击复制】
【打印本页】   【下载PDF全文】   【查看/发表评论】  【下载PDF阅读器】  
←前一篇|后一篇→ 过刊浏览
分享到: 微信 更多
一种基于改进SSD网络的猪个体目标检测方法研究
黄 静, 张 健
(浙江理工大学信息学院,浙江 杭州 310018)
syhj_sy@163.com; gxzj126@126.com
摘 要: 猪个体的目标检测对实现猪养殖过程的精细化管理,促进猪养殖业的智能化与信息化升级具有重要意义。针对实际猪舍环境光照较暗,猪个体被遮挡导致检测困难的实际问题,同时兼顾实时性监测的需求,提出一种基于SSD(Single Shot MultiBox Detector)网络改进的猪个体目标检测算法。首先利用大量猪个体图像对改进后的SSD网络进行训练,然后对猪个体目标检测网络进行评估,最后使用真实的猪舍视频对该网络模型进行测试。结果显示,改进后的SSD网络在对猪个体进行检测时有96.38%的平均精度,在实际具有光线变化和目标遮挡情况的视频中平均能达到21.6 FPS。通过对图像帧的随机抽样发现,漏检率降低了3.65%,误检率降低了1.45%。
关键词: 目标检测;SSD;特征金字塔;注意力机制
中图分类号: TP391.4    文献标识码: A
基金项目: 浙江省重点研发计划项目(2021C01048).
Research on a Target Detection Algorithm of Individual Pig based on Improved SSD Network
HUANG Jing, ZHANG Jian
(School of Information Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China)
syhj_sy@163.com; gxzj126@126.com
Abstract: Target detection of individual pig is of great significance to realizing the fine management of pig breeding process and promote the intelligent and information upgrading of pig breeding industry. Aiming at the detection difficulty of individual pig blocking in the dark pig house, this paper proposes an improved pig target detection algorithm based on SSD (Single Shot MultiBox Detector) network, taking into account the needs of real-time monitoring. Firstly, a large number of individual pig images are used to train the improved SSD network; then, the individual pig target detection network is evaluated; finally, the network model is tested with real pig house video. Results show that the improved SSD network has an average accuracy of 96.38% in the detection of individual pigs, reaching an average of 21.6 FPS (Frames Per Second) in the actual video with illumination change and target occlusion. Through random sampling of image frames, it is found that the missed detection rate is reduced by 3.65%, and the false detection rate is reduced by 1.45%.
Keywords: target detection; SSD; feature pyramid; attention mechanism


版权所有:软件工程杂志社
地址:辽宁省沈阳市浑南区新秀街2号 邮政编码:110179
电话:0411-84767887 传真:0411-84835089 Email:semagazine@neusoft.edu.cn
备案号:辽ICP备17007376号-1
技术支持:北京勤云科技发展有限公司

用微信扫一扫

用微信扫一扫