Industrial-grade robust robot vision for screw detection and removal under uneven conditions

As the number of home appliances is expected to continue increasing, despite the decreasing labor force in Japan, there is a need to automate disassembly processes at recycling plants. However, the automated disassembly of outdoor air conditioner units remains a significant challenge owing to unit size variations and exposure to dirt and rust. To address these challenges, this paper proposes an automated system integrating a task-specific two-stage detection method with a lattice-based local calibration strategy. This approach achieves a screw detection recall rate of 99.8% in the presence of severe degradation and ensures a manipulation accuracy of ±0.75 mm without relying on pre-programmed coordinates. In a real-world validation on 120 units, the system achieved a disassembly success rate of 78.3% with an average cycle time of 193 s, confirming its feasibility for industrial applications.