Application of Undecimated Wavelet Transform to Surface Defect Detection of Hot Rolled Steel Plates

Characteristics of surface defects of hot rolled steel plates were analyzed, and an algorithm for detection of hot rolled steel plates was developed. Based on undecimated wavelet transform and mathematical morphology, the algorithm can solve the problem of false alarms by scales and water marks. Positions of defects were determined by modular maximum of inter-scale correlation of wavelet coefficient at first, and then the defects were classified with priori knowledge about characteristic of surface defects on hot rolled steel plates. Experimental results shows that the algorithm is not sensitive to disturbance of scales, water marks and uneven illumination, and the recognition rate of defects is up to 90.23%.