北京航空航天大学学报 ›› 2020, Vol. 46 ›› Issue (8): 1610-1617.doi: 10.13700/j.bh.1001-5965.2019.0531

• 论文 • 上一篇    下一篇

一种可变锚框候选区域网络的目标检测方法

李承昊1, 茹乐2, 何林远1, 迟文升2,3   

  1. 1. 空军工程大学 航空工程学院, 西安 710038;
    2. 空军工程大学 装备管理与无人机工程学院, 西安 710038;
    3. 西安电子科技大学 机电工程学院, 西安 710126
  • 收稿日期:2019-09-29 发布日期:2020-08-27
  • 通讯作者: 茹乐 E-mail:ru-le@163.com
  • 作者简介:李承昊 男,硕士研究生。主要研究方向:计算机视觉、航空电子信息及其智能化。
    茹乐 男,博士,教授,博士生导师。主要研究方向:抗干扰通信、航空电子信息及其智能化。
    何林远 男,博士,副教授。主要研究方向:图像去雾、超分辨率重建、模式识别。
  • 基金资助:
    国家自然科学基金(61701524);中国博士后科学基金(2019M653742);陕西省自然科学基础研究计划(2017JM6071)

Target detection method for region proposal network with variable anchor box

LI Chenghao1, RU Le2, HE Linyuan1, CHI Wensheng2,3   

  1. 1. Aeronautics Engineering Institute, Air Force Engineering University, Xi'an 710038, China;
    2. Equipment Management and Unmanned Aerial Vehicle Engineering College, Air Force Engineering University, Xi'an 710038, China;
    3. School of Mechano-Electronic Engineering, Xidian University, Xi'an 710126, China
  • Received:2019-09-29 Published:2020-08-27
  • Supported by:
    National Natural Science Foundation of China (61701524); China Postdoctoral Science Foundation (2019M653742); Natural Science Basic Research Program of Shaanxi (2017JM6071)

摘要: 目标检测作为计算机视觉领域的热点问题,目前基于深度学习的目标检测方法可以分为2类:两步检测和一步检测,前者有着较高准确性,后者有着较好速度,但是为提高检测的性能两者都引入了锚机制。为提高目标检测系统的性能,基于深度卷积神经网络的两步检测算法引入了注意力引导(AG)模块,通过对候选区域网络(RPN)的锚机制进行引导,使得对于预选锚框形状的选择更具有多样性;同时针对传统的后处理方式非极大值抑制(NMS)算法存在的误检和漏检的问题,提出了一种置信度因子的NMS (Cf-NMS)算法,对于模型的整体性能有着很大的贡献。实验结果说明,所提方法虽然在速度性能上有略微的下降,但是无论是在RPN变体还是现有的先进算法在准确性方面都有提升。

关键词: 计算机视觉, 目标检测, 深度学习, 候选区域网络(RPN), 非极大值抑制(NMS)

Abstract: Object detection is a hot topic in the field of computer vision. At present, object detection methods based on deep learning can be divided into two categories: two-stage detection and one-stage detection. The former has higher accuracy, while the latter has better speed. In order to improve the performance of detection, anchor mechanism is introduced in both categories. In this paper, Attention Guidance (AG) module is introduced in the two-stage detection method based on the deep convolutional neural network, which guides the anchor mechanism of Region Proposal Network (RPN), making the selection of preselected box shape more diversified. At the same time, to solve the problem of false detection and missed detection in the traditional post-processing Non-Maximum Suppression (NMS) algorithm, a Confidence factor NMS (Cf-NMS) method is proposed, which makes a great contribution to the overall performance of the model. Experiment results showed that, although it has a slight decrease in speed performance, the proposed method has an improvement in accuracy in both the RPN variant and the existing advanced method.

Key words: computer vision, object detection, deep learning, Region Proposal Network (RPN), Non-Maximum Suppression (NMS)

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发