Abstract:

Surface defect recognition of metal parts is a hot topic in the field of pattern recognition. Efficient and reliable surface defect identification method can effectively improve production efficiency and maintain production safety. To solve this problem, a surface defect recognition algorithm based on radial basis function (RBF) neural network and particle swarm optimization (PSO) algorithm was proposed. The weight parameters of RBF neural network were determined and improved by PSO algorithm, and the inertia weight of PSO algorithm was processed linearly. The local oscillation of the optimal solution in PSO algorithm was effectively eliminated. The RBF-PSO surface defect recognition algorithm was trained by network aiming at several common defects on the surface of metal parts, and the corresponding actual test was carried out. The recognition accuracy of RBF-PSO surface recognition algorithm proposed in this paper could reach 96%. Compared with traditional neural network algorithm, it had obvious performance improvement.

Key words: radial basis, particle swarm optimization, optimization algorithm, surface defect recognition