Defect Recognition of Welding Image Based on Rough-Fuzzy Network

To solve the problem that defect characteristics extracted from welding images are seriously disturbed by noises and lower precision of recognition algorithm,a defect recognition algorithm based on rough fuzzy neuron networks is developed.The algorithm takes full advantages of reduction of attributes of rough sets, uncertain data processing of fuzzy sets,self-learning of neuron networks,and approximation to any function.It can solve the problems in the course of uncertainly modeling,such as noise disturbance,(difficulty) in determining model structure.The simulation result shows that the algorithm can effectively improve the recognition of welding images.