针对电瓶车骑手头盔佩戴情况的无人机巡检方法研究

Research on the UAV Inspection Method for the Helmet Wearing State of Electric Bicycle Riders

随着新交通法的实施,电瓶车骑手的头盔佩戴问题引起社会关注,尽管通过使用摄像机可以对骑手进行间接观测,从而减少了直接观察所带来的人力成本,但骑手位置的不确定性,不安全行为的随机性,都给通过固定式摄像机的骑手头盔佩戴检测的准确性和通用性带来了极大挑战。为此,本文设计了一种基于无人机巡检和深度学习的电瓶车骑手佩戴检测方法,该方法首先基于YOLO模型针对无人机采集图片进行分析处理,获取骑手头盔佩戴特征,通过特征分析建立未佩戴头盔骑手的分类模型;然后,针对实际采集过程中的噪声影响,对算法的性能进行验证,并通过噪声抑制方法进一步提升了识别的准确性。所提出的方法可以为电瓶车骑手的安全行驶检测提供新的技术手段,有利于进一步提升交管部门的管理效率。

With the implementation of the new traffic law,the helmet wearing problem of electric bicycle riders has attracted social attention..Although indirect observation of riders can be made through the use of cameras,thus reducing the labor costs associated with direct observation,the diversity of helmet wearing brings a great challenge to the accuracy and versatility of rider helmet wearing detection through fixed cameras..To this end,this paper designs an electric bike rider wearing detection method based on deep learning,which first analyzes and processes the images based on the YOLO model for UAV acquisition,obtains the rider helmet wearing features,and establishes a classification model for riders not wearing helmets through feature analysis;then,the performance of the algorithm is verified against the noise impact in the actual acquisition process,and the noise The accuracy of recognition is further improved by the noise suppression method..The proposed model can provide a new method for the safe driving detection of electric bicycle riders to further improving the management efficiency of traffic control departments.