The second generation of the skeleton and neural network based flexible inspection method for identifying surface flaws

A refined inspection technology which combines neural networks and the range maps of object´s sub-skeleton pixel counts for identifying surface flaws is introduced. The proposed flexible inspection method is a low cost approach and is invariant to object´s position and orientation. The inspection system first performs off-line neural network training and constructs the sub-skeleton range maps merely using an object sample image. The second stage tests flaws on-line based on the combination of the associated neural network classifications and the sub-skeleton range matching. Experimental results demonstrate the feasibility of such an inspection approach and the improvement this presents over the parent work