Lightweight algorithm for subway platform gap foreign object detection based on improved YOLOv7
In the peak period of passenger transport,accidents evolving people or objects getting caught in the gap between the platform screen door and the train door often occur.To address the problem that existing foreign object intrusion detection algorithms have a large number of parameters and calculations,making them difficult to deploy on edge devices with limited computing power and memory resources,a lightweight algorithm model YOLOv7-MobileNetv3 based on improved YOLOv7 was proposed.The backbone network of YOLOv7 was replaced with the MobileNetv3-Large network,resulting in reduced model size,parameter count and improved inference speed.An experimental platform was built based on the subway platform environment,to simulate foreign object intrusion.A subway platform foreign object intrusion dataset was established to train the YOLOv7-MobileNetv3 model.The experimental results show that compared with the original YOLOv7 model,the mean average precision of the improved model is improved by 2.4 percentage points.The detection speed is increased by 13.32 frame/s,and model parameters and computational load are decreased by 32.03％,60.79％.The improved model is more lightweight while ensuring real-time and accuracy,and can be deployed on edge device with limited hardware resources for the detection of foreign objects in subway platforms gaps.
万方数据
维普
