Sobel Filtered UNet Prototype based Date Fruit Classification

Accurate and effective fruit classification is essential for separating and analyzing in the field of agriculture. This article design a proposed model for classifying date fruit using Sobel Filtered UNet (SFU-Net). To increase feature extraction and classification accuracy, the recommended strategy incorporates Sobel edge detection into the UNet architecture. For execution, 1200 images from the KAGGLE Date Fruit Dataset were used. The dataset contains 200 images for six classes of Date fruit for Sugaey, Shaishe, Ajwa, Galaxy, Natab Ali and Rutab date images. Date Fruit Images are preprocessed using the Sobel filter, which is renowned for its ability to identify edges. This Sobel filter highlights the characteristics that are essential for differentiating between various types of date fruit. Subsequently, the preprocessed Sobel filtered date fruit image input is fed into the UNet, a popular CNN that is popular for its efficiency in segmentation tasks. The prototype accomplishes precise date fruit categorization and high-resolution segmentation by using the encoder-decoder structure of UNet. Experiments show that integrating edge detection with state-of-the-art deep learning approaches can lead to notable gains in classification performance when compared to conventional methods. This method not only pushes the boundaries of fruit categorization but also has potential uses in automated sorting methods and agricultural quality control. The results show that the proposed SFU-Net model performs exceptionally well towards date fruit classification, with a high accuracy of 99.45%.