Detection and recognition of objects on images based on MKV-classifiers

• Main Author: Murygin, K.V.
• Format: Article
• Language: English
• Published: Інститут проблем штучного інтелекту МОН України та НАН України 2013
• Series: Искусственный интеллект
• Subjects: Нейронные сети и нейросетевые технологии. Информационная безопасность ИС
• Online Access: http://dspace.nbuv.gov.ua/handle/123456789/85212

In article the algorithm of combination of the binary properties widely used in practice at system engineering of the automatic analysis of the visual information, in the form of the MKV-classifier is offered. Problems of training and using of MKV- classifiers for the decision of detection problems and recognition of objects are considered. The offered algorithms of training allow to generate more effective recognizing rules in comparison with known algorithm AdaBoost, in particular it is essential to reduce number of used properties at identical classifying ability, at the expense of more exact description of position of objects in feature space. Possibility of representation of the MKV- classifier in the form of a decisions tree allows increasing essentially of computing efficiency of classification process.