Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

IMAGE CLASSIFICATION USING SEGMENTATION ALGORITHMS

Dmitriev D.V. 1 Kapranov S.N. 1
1 Nizhny Novgorod State Technical University n.a. R.E. Alekseev
The paper considers a machine-learning problem for image classification and detection on the scene. The “Bag of words” approach has used for image classification. The article involves the following classificatiors: SVM, DTree, RTree, ERTree, GBTtree, Boost, Bayes, KNearest. The original classification algorithm was extended with segmentation and filtration steps. On the segmentation step the object key points were separated from background and another objects on the scene using GrabCut and Watershed algorithms. The filtration step used for object key points allocation. This object should be detected on the image. The algorithm was implemented using computer vision library OpenCV. These algorithm improvements allow to decrease classification errors about 20–30 %.