基于ViBe和目标检测的视频烟雾检测算法  

Video Smoke Detection Algorithm Based on ViBe and Target Detection

作　　者：王林[1] 薛国阳 Wang Lin;Xue Guoyang(School of Automationand Information Engineering,Xi’an University of Technology,Xi’an 710048,China)

机构地区：[1]西安理工大学自动化与信息工程学院,西安710048

出　　处：《单片机与嵌入式系统应用》2023年第4期33-37,共5页Microcontrollers & Embedded Systems

摘　　要：提出了一种基于ViBe与目标检测的VSmokeNet烟雾检测算法。首先,该算法通过改进ViBe算法,减少噪声点的引入,提取到完整的运动前景区域,完成对烟雾的粗筛选;接着,利用大样本数据集训练的改进YOLOv5s网络对粗筛得到的烟雾运动前景区域进行二次筛选,最终实现视频烟雾的精确框定。实验结果表明,该算法在各种场景下有良好的检测效果。In the paper,a VSmokeNet smoke detection algorithm based on ViBe and target detection is proposed.Firstly the algorithm improves the ViBe algorithm to reduce the introduction of noise points,extracts the complete moving foreground region,and completes the rough screening of smoke.Then,the improved YOLOv5s network trained by large sample data sets is used to screen the smoke motion foreground area obtained by coarse screening for the second time,and finally the accurate framing of video smoke is achieved.The experimental results show that the algorithm has good detection effect in various scenes.

关 键 词：目标检测 烟雾检测 ViBe YOLOv5s 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]