Research on abnormal state detection technology of gas station based on video monitoring

In order to meet the requirements of security monitoring and stable operation in key regions of gas station, this paper proposes an algorithm framework for abnormal state detection and recognition of gas station based on video surveillance. Firstly, based on the collected video data of the station, the key frame extraction method of dynamic and static feature fusion is studied, the image key frame are preprocessed by random rotation, scale change and color dithering, and the image data expansion model of the station is established. At the same time, Real-ESRGAN model is further used to achieve super resolution enhancement for weak quality images in the data set, which provides data support for the subsequent detection and recognition algorithm model. At the same time, the attention mechanism is introduced into the YOLO-tiny4 algorithm, combined with the multi-target tracking of personnel and the behavior decision rules, to realize the identification of the multi-target personnel and the travel path of the station, and at the same time to detect the regional key objects and security perimeter. The algorithm model constructed in this paper is verified on the self-built data set, and the test effect is improved to some extent compared with the original algorithm.