Detection and Recognition of Steel Ball Surface Defect Based on MATLAB

Abstract

Taking MATLAB as the experimental platform, the detection and recongnition method of steel ball surface defect is put forward. With the method, we could determine whether there are surface defects or not, and identify the types of defects. The process of detection and recognition is as follows: Firstly, two-time wavelet de-noising treatment of the steel ball image is achieved by means of ecomposition, quantization of threshold and recongnition of Sym4 wavelet function, afterward, many collected noise of the steel ball image is reduced effectively. Secondly, the de-noised image is preprocessed and then we can calibrate the boundary of the defects accurately, which is used to extract the characteristic parameters of defects. Thirdly, the types of defects of steel ball are judged, and the process of pattern recognition is reasonablely designed by putting forward the shape parameter F, combined with the characteristic parameters of the defects. Lastly, the feasibility and validity of the detection and recongnition algorithm are verified by lots of analysis about experimental results especially the analysis on the experimental results comparing with the given data.

DEEP convolutional neural network of microscopy images for pearl defects detection

“…Negate and binarize the resulting image after edge detection [12] . Finally use the software to calculate the proportion of the defect area [13] .…”