基于深度学习的电力作业现场风险识别技术

Risk Identification Technology for Power Operation Site Based on Deep Learning

在带电高空环境作业时,通常要求作业人员佩戴安全帽、安全绳等防护措施,同时规范作业人员操作行为,避免触电等安全事故发生,一旦发现有安全隐患要及时发出警告。传统的监督方式为人工检测,通过现场安全员或者是监控摄像头的监督,存在人工成本高、效率低下的问题。随着深度学习在图像处理领域的发展,实时的目标检测与测距技术已经越来越多地应用在电力作业的安全防控领域。但目前的检测算法针对电力作业环境存在外部干扰大、检测准确率低的问题。因此本文在最新的YOLO v8的基础上,引入了BoTNet模块优化现有目标检测算法,并且仅在出现作业人员的区域进行目标检测,有效减少了检测时间,提高了目标检测的正确率。

When working in an electrified high-altitude environment, it is usually required that the operators wear protective measures such as safety helmets and safety ropes,and regulate their operating behavior to avoid safety accidents such as electric shock.Once a safety hazard is discovered,a warning should be issued in a timely manner.The traditional supervision method is manual inspection,supervised by on-site safety officers or surveillance cameras, which all have the problems of high labor costs and low efficiency. With the development of deep learning in the field of image processing, real-time object detection and ranging technology has been increasingly applied in the safety prevention and control of power operations.However,current detection algorithms have problems with high external interference and low detection accuracy in the power operation environment.Therefore,based on the latest YOLO v8,this article introduces the BoTNet module to optimize existing object detection algorithms,and only performs object detection in areas where there are operators, effectively reducing detection time and improving the accuracy of object detection.