视觉显著物体检测综述:方法、挑战及未来  

Review of salient object detection:methods,challenges and directions

作　　者：刘铁 陈楠 张瀚丹 尚媛园[1] 丁辉[1] 邵珠宏[1] LIU Tie;CHEN Nan;ZHANG Handan;SHANG Yuanyuan;DING Hui;SHAO Zhuhong(Information Engineering College,Capital Normal University,Beijing 100048)

机构地区：[1]首都师范大学信息工程学院,北京100048

出　　处：《首都师范大学学报（自然科学版）》2024年第6期36-48,共13页Journal of Capital Normal University：Natural Science Edition

基　　金：北京市自然科技基金面上项目(4242034)。

摘　　要：视觉显著物体检测作为计算机视觉领域的关键研究方向,也是学术研究的热点之一。本文系统性地梳理了该领域的研究方法、面临的挑战和未来的发展方向。首先,概述了视觉显著物体检测的发展脉络,以及其在计算机视觉领域的广泛应用;其次,对视觉显著物体检测方法进行了详尽的回顾,涵盖了基于显著性特征以及深度学习框架下的检测方法;再次,深入探讨了基于传统卷积神经网络和全卷积神经网络的显著物体检测方法,以及基于注意力机制的显著物体检测方法,并对视觉显著物体检测领域常用的数据集和评价指标进行了介绍;从次,针对当前视觉显著物体检测面临的挑战,如现有数据集的局限、复杂场景下的检测准确度等,文章进行了总结分析;最后,展望了视觉显著物体检测的未来发展方向。通过本文的综述,旨在为从事视觉显著物体检测的研究者提供全面而深入的参考,以促进该领域的进一步发展。Salient object detection,as a key research direction in the field of computer vision,is also one of the hotspots of academic research.In this paper,we systematically sort out the research methods,challenges and future development directions in this field.First,the development of salient object detection is summarized,as well as its wide application in the field of computer vision.Second,a detailed review of salient object detection methods is provided,covering detection methods based on saliency features as well as those under the deep learning framework.Third,salient object detection methods based on traditional convolutional neural networks and full convolutional neural networks,as well as salient object detection methods based on the attention mechanism,are discussed in depth,and commonly used datasets and evaluation metrics in the field of salient object detection are introduced.Again,the article summarizes and analyzes the current challenges of salient object detection,such as the limitations of existing datasets and the detection accuracy in complex scenes.Finally,it looks forward to the future development direction of salient object detection.Through this review,this article aims to provide a comprehensive and in-depth reference for researchers engaged in salient object detection in order to promote the further development of this field.

关 键 词：显著物体检测 计算机视觉 深度学习 注意力机制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]