Computer Engineering and Applications ›› 2023, Vol. 59 ›› Issue (18): 1-13.DOI: 10.3778/j.issn.1002-8331.2305-0310

• Research Hotspots and Reviews • Previous Articles     Next Articles

Review on Research and Application of Deep Learning-Based Target Detection Algorithms

ZHANG Yangting, HUANG Deqi, WANG Dongwei, HE Jiajia   

  1. School of Electrical Engineering, Xinjiang University, Urumqi 830017, China
  • Online:2023-09-15 Published:2023-09-15



  1. 新疆大学 电气工程学院,乌鲁木齐 830017

Abstract: With the continuous development of deep learning, deep convolutional neural networks are increasingly used in the field of target detection and are now applied in many fields such as agriculture, transportation, and medicine. Compared with traditional feature-based manual methods, deep learning-based target detection methods can learn both low-level and high-level image features with better detection accuracy and generalization ability. To outline and summarize the latest advances and technologies in the field of target detection, the status of deep learning-based target detection algorithms and applications is reviewed by analyzing the deep learning-based target detection technologies in recent years. Firstly, the development, advantages and disadvantages of two kinds of target detection network architectures, two phases and single phase, are summarized; secondly, the backbone network, data set and evaluation metrics are described, the detection accuracy of classical algorithms are compared, and the improvement strategies of classical target detection algorithms are summarized; finally, the current stage of target detection applications are discussed, and future research priorities in the field of target detection are proposed.

Key words: target detection, deep learning, computer vision, deep convolutional neural network

摘要: 随着深度学习的不断发展,深度卷积神经网络在目标检测领域中的应用愈加广泛,现已被应用于农业、交通和医学等众多领域。与基于特征的传统手工方法相比,基于深度学习的目标检测方法可以学习低级和高级图像特征,有更好的检测精度和泛化能力。为了概括和总结目标检测领域的最新进展和技术,通过分析近年来基于深度学习的目标检测技术,对基于深度学习的目标检测算法与应用现状进行综述。归纳了两阶段与单阶段两种目标检测网络架构的发展及优缺点;从骨干网络、数据集和评价指标等方面进行叙述,对比了经典算法的检测精度,总结经典目标检测算法的改进策略;讨论了现阶段目标检测应用,并提出了目标检测领域今后的研究重点。

关键词: 目标检测, 深度学习, 计算机视觉, 深度卷积神经网络