摘要
在电力场景下,由于光线变化、相似度高,导致传统算法准确率较低。针对这种情况,提出一种小样本目标检测方法,采用迁移学习的精简模型自动提取物品的特征,可以克服环境变化的干扰,能准确地识别目标,为实现电力场景下作业机器人智能作业奠定基础。实验证明相对于传统方法,本文方法在保证识别速度的同时准确率可提升8%。
In electricity scenes,because of light change,similarity,the recognition rate of the object is very low through traditional algorithms.In or⁃der to overcome the interference of complex background,a method of object detection based on Deep Convolutional Neural Network is pro⁃posed,which extracts the characteristics of object.A simplified model by a transfer learning is used to identify targets efficiently and accu⁃rately in the image to achieve the automatic working.Experiments show that the accuracy of this method can be increased by 8%while achieve satisfied detection and recognition speed in real scenes.
作者
李晨曦
娄根
李慧姝
方武
LI Chengxi;LOU Geng;LI Huishu;FANG Wu(Information Department,Suzhou Institute of Trade&Commerce,Suzhou 215009;Jiangsu R&D Center of Intelligent Service Engineering Technology,Suzhou 215009)
出处
《现代计算机》
2021年第18期109-112,共4页
Modern Computer
基金
2019年教育部科技司-赛尔网络“下一代互联网技术创新项目”(No.NGII20190701)
2020年苏州经贸职业技术学院院级项目(No.YJ-ZK2012)。
关键词
目标检测
卷积神经网络
数据增强
Object Detection
Deep Convolutional Neural Network
Data Argumentation
