Visual inspection system for smart manufacture of home appliances

Home appliance manufacturing involves a lot of human operations due to variety of products and sizes, and lower cost requirements. Applying traditional inspection solutions to home appliances faces three challenging issues: 1)hard to handle different sizes of objects from big refrigerators to small kettles with fixed camera setting; 2)heavy engineering and tuning work for different inspection tasks; 3)high cost of stand-alone inspection system. This paper proposes a camera mounted robot solution with self-learning and edge/cloud visual inspection architecture. Firstly the 4 DoF robot arm with novel controlling strategy uses a single camera to handle variety of objects. For different inspection tasks, a machine learning based approach removes the complicated and tedious works for engineering and parameter tuning. Lastly, an edge/cloud computing framework puts limited tasks on cheap edge devices at manufacturing side and reduces the total cost.