| Abstract [eng] |
This work aims at creating the rational medical anesthetic masks’ defect detection algorithms for the computer vision system. Definitions of the defects were obtained from the customer. The user interface of the computer image analysis system should allow an easy access to the filter settings. The device could be installed in a real production line, where it could work synchronously with a robot. In the first part of this study the image analysis methods are examined. As well the analytical review of similar systems is presented. In the second part the masks’ defects are investigated. In addition, the searching for the rational algorithms of their identification is carried out. A series of algorithms are created for the each type of the defect; such algorithms are assessed according to the following criteria - accuracy, reliable and fast performance. The experimental studies of all the possible algorithms are carried out. It was determined the necessary number of images, position of masks in order the larger surface area during the analysis could be inspected. The filters are designed for a structured list of defects in MS Visual Studio programming tool environment. The convenience of the programme user’s interface is assessed. The required parameters are included in the user’s interface, which can be changed in future if there is a necessary to adjust the system or in case of changes in the product (size, type). The third part of this study provides the system implementation. The criteria of the system hardware choice are analysed. As well the final structure of the device block diagram is presented. The speed of the computer vision machine, according to the number of images and engine parameters, is assessed. The reliability of the system in real production conditions is observed. |
