# Research and Application of Safety Hazard Perception and Responsibility Traceability System in University Laboratories

## TL;DR

This paper introduces a deep learning-based system to enhance university lab safety by detecting hazards and tracing responsibilities.

## Contribution

A novel system integrating object detection, attention mechanisms, and tracking for safety hazard perception and responsibility traceability in university labs.

## Key findings

- The system achieved 87.8% mean average precision in laboratory hazard detection.
- Multi-target tracking effectively identified and traced unsafe behaviors like not wearing a lab coat.
- The system generates traceability reports, solving issues of visibility without control in traditional lab safety management.

## Abstract

In order to solve the challenges of laboratory safety management in universities, such as insufficient supervision of high-frequency risk behaviors in responsibility traceability, a laboratory safety hazard perception and responsibility traceability system based on deep learning is proposed. Based on the YOLOv5s object detection model, the channel attention mechanism SE and NWD loss functions are introduced, with DeepSORT tracking to realize multi-target tracking and hidden danger perception in laboratory scenarios. Then, the responsibility matching algorithm and visual traceability mechanism are proposed to build a full-chain management system of “risk perception, analysis and tracking, and responsibility traceability”. Experiments show that the mean average precision (mAP) of YOLO-lab in the laboratory scene is 87.8%. Taking the experimenter not wearing a lab coat as an example, through the test of the laboratory scene, the multi-target tracking effect is excellent and the responsibility traceability report is generated, which solves the problem of “visible and uncontrollable behavior, traceable and unproven” in traditional supervision, and provides an intelligent technical path for laboratory safety governance.