Abstract
In on-line inspection for products based on machine vision, contrast between product defects and background is not always stable because of the discrepancy of environment where images are captured and the type, batch and innate structure of products to be detected. To perform accurate detection, image is usually divided into several parts which are of same gray value and later on sub-blocks the analysis for defected region where sudden gray value changes are occurring. The crucial step here is to have accurate regional image segmentation. Traditional edge detection is unlikely to ensure its accuracy, and at the same time, complicated image segmentation algorithms are time-consuming and cannot meet needs of real-time manufacturing. Images captured during on-line detection is relatively stable in structure. A new real-time image fast segmentation algorithm is proposed in this dissertation. This algorithm, combining with use of local image enhancement algorithm, morphological operation of simple structural operators and image thinning technology, can accurately find regions boundry of uniform region. Later, on-line image segmentation can be fulfilled by means of simple addition and subtraction for regions. This algorithm has been successfully applied to on-line capsule inspection. Experiments show that it can satisfy the need of on-line detection both with speed and accuracy.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call