期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2021
卷号：99
期号：11
语种：English
出版社：Journal of Theoretical and Applied
摘要：Quality control using image processing and machine vision offers faster performance and high accuracy, with minimum human contact. A significant application refers to bottle counting in a box-based pattern to deliver appropriate quantity, according to customers' demand. The precise arrangement with very close proximity to each other poses a significant challenge. As a result, missed detection tends to occur during segmentation due to interconnected boundaries and contacts. The watershed algorithm provides a possible solution with the capacity to segregate touching and overlap. A particular approach to bottle detection and counting used in this research. The results showed the average detection accuracy for 9-12 bottles at 99.5726%. However, with mean values for memory usage on every detection of 0.6 MiB and time interval on every 3 boxes detection within 0.21 secs, the use of the watershed algorithm was strongly believed as a suitable alternative.
