Crawler crane overturning often results in many casualties and property damage. However, the existing research on overturning prevention mainly focuses on the internal factors of crawler cranes while ignoring the environmental factors represented by the subgrade bearing capacity. On this basis, this work summarizes three types of dangerous work zones including dangerous work zone with pit, dangerous work zone with unhardened area, and dangerous work zone with water with poor subgrade bearing capacity and develops an automated method for detection. A Mask Transformer model is adopted by using Swin Transformer as backbone network to recognize and segment the images obtained from an unmanned aerial vehicle. The detected images are transformed into a safety risk map that provides the driver with risk information about the dangerous work zone. Results show that the model proposed, which has been applied in a real engineering project, achieves a good detection effect.
Read full abstract