Abstract:

In order to guarantee the characteristics of real time and accuracy in strip steel defect classification, we propose a defect classification model based on  hybrid chromosome for the  strip steel surface image. This method not only optimizes both the kernel function parameters and the penalty factors of the support vector machine (SVM) model, but also makes choice of feature vectors and kernel functions. Meanwhile, the genetic algorithm and the SVM model are fused and the final SVM classifier is established by using the decoding results of the optimal chromosome in the end. Experimental results show that our method is effective and efficient in classifying the defects in the strip steel surface image.

Key words: strip steel surface defect classification;hybrid chromosome;SVM;genetic algorithm;classification accuracy