Visual inspection of glass bottlenecks by multiple-view analysis

Abstract

Automated inspection using multiple views has been recently developed to automatically detect flaws in manufactured objects. The main idea of this strategy is that, unlike the noise that appears randomly in images, only the flaws remain stable in a sequence of images because they retain their position relative to the movement of the object being analyzed. In this paper the design of a prototype for image sequence acquisition and inspection of glass bottlenecks using this methodology is proposed. The inspection device considers light sources inside the bottle for better capture of potential defects. Using a single camera we record an image sequence of the bottleneck rotating the bottle along its principal axis. No camera calibration process is considered. We also present a novel approach for defect detection based on tracking potential flaws along multiple views. Our inspection system achieves over a 98% of true positives and 2.4% of false positives rates under uncalibrated image sequences. These results allow us to conclude that uncalibrated images sequences can be used effectively as a mean to evaluate the product quality where the images of the test object do not significantly differ from each other in the sequence.