Aiming at the problems that the two parameters of BTSVM optimized by weld surface defect PSO are easy to fall into local optimum,low pseudo-random diversity and low calculation accu-racy,a method of LCSPSO to optimize BTSVM parameters is proposed.Firstly,improve the PSO algo-rithm,introduce Levi's flight strategy,Circle chaotic map and simplex reflection operation to ensure the global optimal solution of the two parameters and improve the classification accuracy.Reasonable classi-fier uses the improved method to classify and detect weld surface defects and compares it with other al-gorithms.The test results show that LCSPSO is superior to other algorithms,the classification accura-cy is 12.03％higher than that before improvement,and the average classification accuracy is 99.07％.
维普
万方数据