Hypothesis-validated construction accident analysis and framework for potential multi-sensor fusion with 360° camera and LiDAR

Construction sites are among the most hazardous work environments, with frequent accidents such as falls, entrapments, and collisions. Traditional single-sensor detection systems suffer from occlusions, poor lighting, and limited depth perception, reducing reliability in complex conditions. This study addresses these limitations by proposing a realtime multi-sensor fusion framework integrating a 360° camera and LiDAR. A three-year construction accident dataset was analyzed, and Chi-squared tests and ANOVA (p < 0.05) confirmed the statistically significant superiority of the fusion approach over single-sensor systems. Deep learning techniques were applied to enhance real-time detection and prevention capabilities. The results demonstrate that sensor fusion substantially improves detection accuracy, especially in high-risk scenarios such as falls and collisions. This study provides a comprehensive statistical analysis based on national incident records and proposes an AI-driven monitoring framework. The findings offer strong empirical support for multi-sensor fusion as a foundation for next-generation construction site safety systems.