基于计算机视觉算法的零部件缺陷智能检测系统

打开文本图片集

中图分类号：U463.6 文献标识码：A 文章编号：1003-8639（2026）01-0111-03

Design and Application of Surface Defect Detection System forAutomobile PartsBased on Computer Vision Recognition

（BaojiUniversityofArtsand Sciences，Baoji721O16，China）

【Abstract】With the transformation of the automobile industry to high-end manufacturing，the existing automatic defectdetection equipment is insuficientindealing with curved workpieces，reflective materials，and compositedefects. Therefore，this paper designs and implements an intellgent inspection system based on computer visionalgorithms： in theimagingpart，the depth offield expansionalgorithm is used torealize theglobalcoverage ofcurvedworkpieces;In image processing， HSVcolor space conversion and adaptive histogram equalization algorithm are introduced to suppress reflectionanduneven ilumination.Indefectdetection，deep learning modeland threshold segmentationalgorithmare integrated to accurately locate small defects.After testing，the detection rate of O.2 mm scratch reaches 98.7% ，and the misjudgment rate of bubbles and oil stains is only 0.3% . The detection time of single piece is stable at 42O milliseconds， and the influence of environmental temperature drift is controlled within 3% error band.

【Key Words】computer vision;defect detection;multi-angle imaging system;threshold segmentation algorith

0 引言

随着汽车生产工艺逐步向精确性、灵活性方向持续迈进，每辆汽车的核心零部件数目高达上万个，检测人员日常需分辨数万组相似纹理，而人眼对小于 0.5mm 缺陷的漏检率较高。（剩余4021字）