基于改进YOLOv7的吸烟行为识别算法研究

Research on smoking behavior recognition algorithms based on improved YOLOv7

目的:提出了一种基于改进YOLOv7的吸烟行为识别方法,以提高人工智能对吸烟行为识别X效率和准确率。方法:在YOLOv7算法的基础上,采用GhostNet网络结构替换其主干网络,减少了网络模型参数量和计算量。引入CBAM注意力机制来改善特征提取的效果。加入多尺度特征融合模块和改进损失函数以提高模型在复杂环境中的检测效果。结果:以吸烟数据集进行测试,结果表明,改进后的模型参数量和计算量分别下降了16.6%和37.4%,检测速度提升至103.4帧/s,精度提高了2.8%。结论:提出的轻量化网络模型可满足视频监控实时监测的要求,并可在低功耗的嵌入式设备上完成实时检测。

Aims:A smoking behavior recognition method based on improved YOLOv7 was proposed to improve the efficiency and accuracy of artificial intelligence in smoking behavior recognition.Methods:Based on the YOLOv7 algorithm,the GhostNet network structure was used to replace its backbone network,reducing the number of network model parameters and computational complexity.The CBAM attention mechanism was introduced to improve the effectiveness of feature extraction.A multi-scale feature fusion module and an improved loss function were incorporated to enhance the model s detection performance in complex environments.Results:Testing conducted on a smoking dataset showed that the improved model reduced the number of parameters and computational complexity by 16.6%and 37.4%,respectively.The detection speed was improved to 103.4 F/s;and the accuracy was improved by 2.8%.Conclusions:The proposed lightweight network model can meet the requirements of real-time video monitoring and can achieve real-time detection on low-power embedded devices.