Classification and Segmentation of Watermelon in Images Obtained by Unmanned Aerial Vehicle

Abstract

In this study, watermelons in the images obtained by an unmanned aerial vehicle (UAV) from watermelon field in Adana, Turkey, were segmented and classified. The original image obtained was processed in two ways. To start with, images were divided into overlapping blocks and gray level co-occurrence matrix (GLCM) from these blocks was generated and texture features were extracted using these GLCMs. Then, blocks containing or not containing watermelon were classified by employing a linear classifier. As a result of this study, the accuracy of watermelon classification was obtained as 86.46%. Second, k-means clustering of the original image was performed. Following this, groups having the highest blue value at its center were chosen. Finally, first and second results were combined with logical and operator. It was derived from this study that the method may be useful for detecting watermelons in field images obtained via UAV for counting and yield estimation.