RGB-D BASED WORK AREA RECOGNITION MODEL FOR FIREPROOFING SPRAY ROBOT

Abstract: Fireproofing spray on steel structures is essential for fire resistance, yet it is still manual, exposing workers to hazardous and dusty conditions. Robots have been developed to address these issues, but some systems depend on work area map, which limits real time adaptation. In addition, object level recognition cannot provide surface level work areas. Therefore, we propose an RGB D work area recognition model that classifies work area surfaces on a steel structure beam. On experiment, the model achieves a macro average F1 score of 0.8682. Finally, the proposed model provides surface level work area for fireproofing spray robot.