Defect Detection Method for Die-casting Aluminum Parts Based on RESNET

Abstract

Surface defect detection of die-cast aluminum parts is always the focus of auto-mobile quality control. Most of the existing algorithms are designed to detect defects in a particular working condition. Although the effect is good, the application scope is relatively narrow. In the field of surface defect detection of die-cast aluminum parts, one of the current challenges is to segment target detection positions from complex field camera images and effectively detect defects in products in real time. In this paper, a defect detection algorithm combining traditional digital image processing algorithm and deep learning algorithm is proposed. The tar-get detection area is cut out timely and effectively through traditional image processing, and then the target area is classified by using residual network. The experimental results on the surface defect data set of die-casting aluminum parts show that the detection speed of this algorithm is very fast, and the accuracy rate reaches 98%.