基于YOLOv7和GMM算法的火焰实时检测方法  

Real-time Flame Detection Method Based on YOLOv7 and GMM Algorithm

作　　者：王振龙 陈彦 刘飞 王优优 杨罗刚 张佳铭 WANG Zhenlong;CHEN Yan;LIU Fei

机构地区：[1]石化盈科信息技术有限责任公司,北京100000

出　　处：《科技创新与应用》2024年第5期1-6,共6页Technology Innovation and Application

摘　　要：石油化工行业生产现场大多具有易燃、易爆、高温和高压等特点,是安全监管的核心区域。针对目前安全监管区域主要依靠监控视频回放及人工检查方式,导致实时性差、安全性低的问题,该文提出一种基于YOLOv7和GMM算法的火焰实时检测方法。首先,通过生产现场对数据实时采集和分析,基于YOLOv7在目标检测领域的优势,实现对疑似火焰区域的检测;其次,结合火焰特有的动态特征,利用GMM模型对检测结果中的疑似火焰区域进行排除,如灯光、太阳光等;最后,利用形态学操作方法,进一步提高对真实火焰检测的准确性。通过实验结果表明,该方法能够滤除疑似火焰的干扰,有效检测出真实火焰,且在复杂背景下仍有很好的抗干扰能力和识别准确率。Most production sites in the petrochemical industry have the characteristics of flammability,explosion,high temperature,and high pressure,and are the core areas of safety supervision.Aiming at the problem that the current safety supervision area mainly relies on surveillance video playback and manual inspection,which leads to poor real-time performance and low security,a real-time flame detection method based on YOLOv7 and GMM algorithm is proposed in this paper.First of all,through the real-time data collection and analysis on the production site,based on the advantages of YOLOv7 in the field of target detection,the suspected flame area is detected;secondly,combined with the unique dynamic characteristics of the flame,the GMM model is used to eliminate the suspected flame area in the detection results,such as light,sunlight and so on;finally,the morphological operation method is used to further improve the accuracy of real flame detection.The experimental results show that this method can filter out the interference of suspected flame,detect the real flame effectively,and still has good anti-jamming ability and recognition accuracy under complex background.

关 键 词：易燃 高温 YOLOv7 GMM 形态学操作 

分 类 号：X932[环境科学与工程—安全科学]