Abstract:

Monitoring products is important for quality and ripening control in an efficient agricultural production process. Monitoring is mostly done with captured images and videos in accordance with the developed technology. The quality of these images and videos directly affects the evaluation. If there is a distortion in image or video, first of all, this distortion must be detected and classified to eliminate. In this study, a method is presented to classify distortions in agricultural images. Eleven different distortions are synthetically added to agricultural images. A convolutional neural network (CNN) is designed to classify distorted images. The designed CNN model is tested with four different datasets obtained from various agricultural fields. Also the designed CNN model is compared with previously presented CNN architectures. The results are evaluated and it is seen that the designed CNN model successfully classifies distortions.

Keywords: