| Abstract [eng] |
Investigation of methods for counting electronic components on a conveyor. Masters’ graduation thesis for electronics engineering degree, Vilnius Gediminas Technical University. Vilnius, 2022, 71 p., 27 pictures., 16 tables., 59 lit., 2 ex. The aim of this thesis is to investigate intelligent techniques used to identify and detect tiny objects and apply chosen method for counting electronic components on a conveyor. For achieving the aim, an analytical review of image processing and classification, deep learning and object detection methods was conducted. An individual dataset for training regression-based object detectors to identify electronic components was collected and a method for counting electronic components was formulated. During the experimental research, trained object detection models were implemented and investigated, their image processing speed, mean average precision, inference time while detecting electronic components were evaluated. After the implementation of the proposed calculation method, quantities of the calculated components and the effect of the method on the efficiency of the models were additionally evaluated and compared. The system is capable to identify and count four different types of electronic components. |
