Smoking scene recognition method based on super-resolution and time sequence analysis

The invention discloses a smoking scene recognition method based on super-resolution and time sequence analysis, and belongs to the field of intelligent monitoring, intelligent security and image recognition and processing. According to the method, the MobileNetV3-CBAM is used for preliminarily recognizing the behavior of a person, a clear picture is reconstructed through super-resolution, then, AlphaPose is used for behavior logic judgment, and finally, a video recognition model is used for completing time sequence analysis and accurate recognition. According to the technology, restriction and influence of long-distance and complex scenes on smoking behavior detection can be effectively weakened in actual production. The technology is helpful for guaranteeing the safety of the production environment and the seriousness of the law enforcement scene, and plays an important role in the current social background. 本发明公开了一种基于超分辨率和时间序列分析的抽烟场景识别方法，属于智能监控、智能安防、图像识别及处理领域。本发明使用MobileNetV3-CBAM初步识别人物行为，经过超分辨率重建清晰图片，再利用A