Multilayer local binary patterns for grape types detection

Grapes with a large number of varieties are among the most important agricultural products grown all over the world. Observation and classification of grape species requires very robust botanical knowledge. This has often been a very difficult area for botanical scientists too. In addition, the reduction in the number of botanists has made it necessary to use different methods for identification of plant species, such as techniques based on image processing and identification of relationships. In addition, developments in different areas related to the subject, such as the development of camera systems and other portable devices, have increased the applicability of this idea. The identification of grapes and the correct identification of grape varieties are of systematic importance. In this study, a computer vision-based method was proposed as an alternative to classical diagnostic methods for the detection of grape varieties. Grape seed images with characteristic features were used for the classification of grape species. In the study, A new approach based on Local Binary Pattern (LBP) has been proposed to produce effective features from color kernel images. Recommended method is used for color images. The detection of grape varieties was carried out in two stages. Attributes were obtained in the first stage, and in the second stage, classification by 10-fold cross validity test was performed by using different machine learning methods through these attributes. At the end of the study, the classification success was obtained as 82,10%.