计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (14): 16-26.DOI: 10.3778/j.issn.1002-8331.2203-0506

• 热点与综述 • 上一篇    下一篇

林火视频烟雾检测算法综述

朱弥雪,刘志强,张旭,李文静,苏佳新   

  1. 1.内蒙古工业大学 信息工程学院,呼和浩特 010080
    2.内蒙古建筑职业技术学院,呼和浩特 010020
  • 出版日期:2022-07-15 发布日期:2022-07-15

Review of Research on Video-Based Smoke Detection Algorithms

ZHU Mixue, LIU Zhiqiang, ZHANG Xu, LI Wenjing, SU Jiaxin   

  1. 1.College of Information Engineering, Inner Mongolia University of Technology, Hohhot 010080, China
    2.Inner Mongolia Technical College of Construction, Hohhot 010020, China
  • Online:2022-07-15 Published:2022-07-15

摘要: 火灾早期时,烟雾相较于火焰特征更为明显,因此早期的烟雾检测对于预防火灾具有重大意义。针对森林、草原火灾早期的烟雾检测,烟雾探测器成本高、检测效果较差且不能提供烟雾大小、方向、位置等信息。随着计算机视觉的发展,基于视频的烟雾检测方法以低成本、覆盖面广、信息获取较全面等优势逐渐成为研究热点。但由于森林、草原背景复杂、烟雾本身易变化,因此视频烟雾检测算法仍面临着巨大的挑战。为研究深度学习的视频烟雾检测中的应用效果,分析了烟雾检测存在的难点问题及传统视频烟雾检测算法的不足,介绍了当前深度学习中各类目标检测算法在烟雾检测中的应用,对比了这些烟雾检测算法,总结了其优点和不足,重点分析了各种烟雾检测难点问题的解决方法,并提出烟雾检测的下一步研究方向。

关键词: 深度学习, 烟雾检测, 特征提取

Abstract: In the early stage of fire, the characteristics of smoke are more obvious than that of flame, therefore, early smoke detection is of great significance for preventing fires. For early smoke detection of forest and grassland fires, smoke detectors have high cost, poor detection effect, and cannot provide information such as smoke size, direction, and location. With the development of computer vision, video-based smoke detection methods are more and more widely used in fire detection due to the advantages of low cost, wide coverage, and comprehensive information acquisition. Therefore, video smoke detection algorithms still face huge challenges. In order to study the application effect of deep learning in video smoke detection, it first analyzes the difficulties in smoke detection and the shortcomings of traditional video smoke detection algorithms, and then introduces the application of various target detection algorithms in current deep learning in smoke detection. Then, these smoke detection algorithms are compared, and their advantages and disadvantages are summarized. Finally, the solutions to various difficult smoke detection problems are analyzed, and the next research direction of smoke detection is proposed.

Key words: deep learning, smoke detection, feature extraction