Blow-Molded Plastic Tube Defect Detection Using a Single Uncalibrated Camera

Authors: Luis Unzueta and Javier Barandiaran and Álvaro Segura

Date: 17.06.2009


PDF

Abstract

In this paper we present computer vision strategies to detect automatically defects on blow-molded plastic tubes used as outer covers of car dampers. A correct quality checking of these tubes is very important as the customers of these pieces are becoming more exigent and, in certain occasions, they reject full sets of thousands of pieces because of some non-valid samples (e.g., a limit of 6 defective tubes per million). The automation of this process is challenging due to the small size of the expected typical defects, the dark color of the molded material and the specular reflection they have. Current standard industrial solutions for this kind of process are not capable of fulfilling the objective. The proposed system uses a laser beam and a single uncalibrated standard camera to scan the tube surface even without the need of any ambient illumination at all. The presented computer vision algorithms reconstruct the tube surface from the observed laser beam deformations and identify the defects by comparing the shape respect to the one expected. Experimental results show the suitability of the proposed methods to detect holes, burrs and deformations on this type of tubes in real time, improving the productivity of the tube quality checking process.

BIB_text

@Article {
author = {Luis Unzueta and Javier Barandiaran and Álvaro Segura},
title = {Blow-Molded Plastic Tube Defect Detection Using a Single Uncalibrated Camera},
pages = {nov-17},
abstract = {
In this paper we present computer vision strategies to detect automatically defects on blow-molded plastic tubes used as outer covers of car dampers. A correct quality checking of these tubes is very important as the customers of these pieces are becoming more exigent and, in certain occasions, they reject full sets of thousands of pieces because of some non-valid samples (e.g., a limit of 6 defective tubes per million). The automation of this process is challenging due to the small size of the expected typical defects, the dark color of the molded material and the specular reflection they have. Current standard industrial solutions for this kind of process are not capable of fulfilling the objective. The proposed system uses a laser beam and a single uncalibrated standard camera to scan the tube surface even without the need of any ambient illumination at all. The presented computer vision algorithms reconstruct the tube surface from the observed laser beam deformations and identify the defects by comparing the shape respect to the one expected. Experimental results show the suitability of the proposed methods to detect holes, burrs and deformations on this type of tubes in real time, improving the productivity of the tube quality checking process.
}
isbn = {978-84-937328-0-6},
date = {2009-06-17},
year = {2009},
}
Vicomtech

Parque Científico y Tecnológico de Gipuzkoa,
Paseo Mikeletegi 57,
20009 Donostia / San Sebastián (Spain)

+(34) 943 309 230

close overlay