Hazard avoidance behaviors (HABs) are critical for preventing injuries in construction, yet remain difficult to assess due to the challenge of capturing whole-body coordination during dynamic tasks such as turning or material handling. This study proposes a more objective method for evaluating HABs by analyzing high-dimensional kinematic patterns using non-intrusive sensing.
Vision-based 3D skeletal recognition extracted whole-body motion data; PCA obtained key movement features, and the entropy weight method quantified behavioral stability and risk. Indoor simulations replicated typical construction scenarios (straight walking and turning) under normal and material-handling conditions.
The analysis revealed clear differences in kinematic patterns and stability between handling and non-handling conditions, as well as between straight walking and turning. The proposed method effectively characterizes subtle variations in movement that are not easily observable, enabling the identification of high-risk behavioral patterns associated with hazard avoidance.
Evaluation was conducted in controlled indoor settings to ensure consistent kinematic pattern analysis. Future studies in more diverse environments and with broader participant groups could extend the method's applicability, and exploring its performance in real construction conditions is a promising research direction.
This study proposes a structured, non-intrusive approach to characterize and quantify HABs as latent unsafe behaviors. Integrating whole-body movement features with objective weighting enhances the understanding of behavioral stability, providing a practical analytical tool for proactive construction safety management.
