Study of clustering methods matching for identification and selection of apples in the conventional line sorting process

Цвіркун, Людмила Олександрівна and Цвиркун, Людмила Александровна and Цвіркун, Сергій Леонідович and Цвиркун, Сергей Леонидович (2019) Study of clustering methods matching for identification and selection of apples in the conventional line sorting process. Обладнання та технології харчових виробництв (1 (38)). pp. 94-101. ISSN 2079-4827

Preview

Text Tsvirkun 16_08_2019+.pdf Download (3MB) | Preview

Abstract

It is noted that the methods and means of algorithmic and informational provision of automated control systems do not allow to identify and select the apples sufficiently effectively in accor¬dance with their characteristics. As a result, it is quite difficult to get high quality control of the process of sorting fruits. The structure of the system for managing the sorting of apples is considered, taking into account the limited speed of the actuator, with the help of which several applets with the best charac¬teristics are selected. The system provides an estimation of the location of objects of selected varieties, taking into account the speed of the executive mechanism, determining the size, color and location of the visual objects in the frame; On the basis of this information, the objects to be selected from the stream of apples, the segmentation of the frame (search of visual objects) is selected. It is emphasized that the affiliation of certain types of apples is determined by the aggregate of characteristics. There¬fore, for the determination of apples to be selected on the conveyor line, it is necessary to investigate clustering methods for the most essential characteristics of apples, since the identification and selection of apples are carried out on the basis of the specified characteristics {d, m, g}. The possibility of apply¬ing clear and fuzzy clustering methods for initial adjustment of the subsystem of the recognition of apple varieties was studied in the formation of the automated control of the sorting of fruits in the stream. The convergence of the clustering algorithms of the characteristics of objects, which was checked by the breakdown into the optimal number of clusters, was analyzed, that allowed to evaluate not only the convergence indices, but also to check the expediency of the choice of the optimal number of clusters previously made. It was found that in order to form an automated control of the sorting of apples in the flow, it is advisable to implement the primary setting of the subsystem of the recognition of varieties us¬ing the Gustafson-Kessel fuzzy clustering algorithm.

Actions (login required)

View Item