摘要
火灾给人们的生命财产安全带来了严重威胁,做好火灾预防有着重要意义。基于人工智能和嵌入式控制技术,开发了一套基于视觉的嵌入式烟火监控云台,用于公共区域的烟火实时监控与预警。该云台使用NumPy算法库对视频图像进行快速预处理,以提高烟火检测的准确度和处理速度;使用Yolov5检测算法对视频图像中的烟雾和火焰进行分类识别,通过选取大量不同的火焰和烟雾场景制作数据集,得到烟火视频图像的检测模型,可以实现大范围的烟火监控与跟踪。实验表明,该算法可以实现通用场景下的烟火快速识别、检测与预警。
Fire has brought a serious threat to people's life and property safety.It is of great significance to do a good job in fire prevention.Based on artificial intelligence and embedded control technology,this paper develops a set of vision based embedded fireworks monitoring PTZ for real-time monitoring and early warning of fireworks in public areas.The PTZ uses NumPy algorithm library to quickly preprocess video images to improve the accuracy and processing speed of fireworks detection.Yolov5 detection algorithm is used to classify and recognize the smoke and flame in the video image.By selecting a large number of different flame and smoke scenes to make data sets,the detection model of fireworks video image is obtained,which can realize a wide range of fireworks monitoring and tracking.Experiments show that the PTZ algorithm can realize the rapid recognition,detection and early warning of fireworks in general scenes.
作者
李博
刘磊
Li Bo;Liu Lei(Tianjin Shengda Fire Industrial Corporation,Tianjin 300381;Tianjin University of Science and Technology,Tianjin 300222)
关键词
云台系统
图像识别
深度学习
烟火监控
PTZ system
image recognition
deep learning
fireworks monitoring
