YOLOv5算法在山火检测中的应用

Application of YOLOv5 Algorithm in Wildfire Detection

随着高新技术的发展,山火的监控、预警方式从过去以人工为主转换成图像、热成像等火焰识别的计算机处理,前者检测成本较高且监视范围小,同时存在环境背景杂乱,灯光、白云等与烟火混淆的因素导致检测误差大,预警延迟等问题。YOLO是基于深度学习的目标检测算法,它对收集到的多元烟火数据集进行以火焰、烟雾为研究对象的目标检测训练,能够得到烟火的自动识别检测模型,同时降低环境中检测混淆因子造成的影响使得分离烟火。对目标检测模型进行优化改进,使最后训练出的检测模型能够对图片或视频有更好的检测效果。

With the development of high and new technology, the monitoring and early warning methods of wildfires have been transformed from manual processing to computer processing of flame identification such as images and thermal imaging. The former has high detection cost and small monitoring range. At the same time, there are some problems such as disorderly environmental background, large detection error and delayed early warning caused by the confusion between lighting, white clouds and fireworks.YOLO is a target detection algorithm based on deep learning. It conducts target detection training on the collected multivariate fireworks data set with flame and smoke as the research objects, and can obtain the automatic identification and detection model of fireworks. At the same time, it reduces the influence of detection confusion factors in the environment to separate fireworks. The target detection model is optimized and improved so that the final trained detection model can have better detection effect on images or videos.