摘要
随着人工智能的发展,视觉感知技术为林业害虫识别防治提供了新方法和实现思路。论文提出一种基于深度感知神经网络框架实现小蠢虫的检测识别,检测系统具体采用了特征金字塔结构、形变结构和新型非极大值抑制技术进行构建,其准确率达到了96.32%。这说明了基于人工智能技术方案识别林业害虫的可行性。与传统方法相比,此方法在精准识别林业害虫的同时,有效减少不必要的资源消耗。
With the development of artificial intelligence,vision perception technology provides new methods and implementation ideas for the control of forest pests.In this paper,a forest pest identification framework based on depth perception neural network is proposed to detect and identify Scolytidae.The detection system contains feature pyramid structure,deformable convolution structure and non maximum suppression,which the pest identification model is 96.32%.Compared with traditional methods,this method can accurately identify forest pests and effectively reduce the consumption of unnecessary resources.
作者
华月珊
王佳新
戎洁庆
华国栋
李莉
HUA Yueshan;WANG Jiaxin;RONG Jieqing;HUA Guodong;LI Li(Guangzhou Chenjing Ecological Technology Service Co.,Ltd,Guangzhou,Guangdong 510520,China;Guangdong Forestry Survey and Planning Institute,Guangzhou,Guangdong 510520,China;School of Information and Electrical Engineering,Hebei University of Engineering,Handan,056038,China)
出处
《林业与环境科学》
2021年第6期124-129,共6页
Forestry and Environmental Science
基金
邯郸市科学技术研究与发展计划项目(1721203049-1)。
关键词
林业病虫害
小蠢虫检测识别
深度神经网络
forest diseases and insect pests
Scolytidae recognition and identification
depth neural network
