Penulis: Asfiyatul Badriyah, Moechammad Sarosa, Rosa Andrie Asmara, Mila Kusumawardani, and Dimas Firmanda Al Riza

Nama Jurnal: BIO Web of Conferences

Tahun: 2024

Volume: 117

Issue: Tidak ada

Halaman: 01046

Deskripsi: Visual identification of diseases in grapevines can be a difficult task for growers. The importance of farmers in the identification of grape diseases due to control the spread of disease and lower agricultural yield losses. In this study developed a disease identification system in plants using image processing. Images of leaves on grapevines infected with the disease were taken, extracted features from the images and applied the ResNet-50 algorithm. The dataset of grape leaf images taken was 200 images for four classes, including 3 classes of leaves identified as diseased and 1 class of healthy leaves. The experimental results show that the image processing system for identifying diseases in grapes identifies the types of disease in grapevines. This research has the potential to be implemented in a farm automation system to detect early diseases in grapevines and take appropriate preventive measures to increase productivity and crop quality.

Pendahuluan: Grapes have a high market value, making them profitable in the economic sector [1]. Additionally, grapes contain antioxidants that are beneficial to human health [2]. Due to the numerous benefits of grapes, the demand for them continues to increase. Identifying plant diseases in grapevines is crucial for improving grape harvests. It is important to do this early and accurately [3]. In addition, grape growth requires just the right amount of sunlight to prevent the grape skin from becoming wrinkled [4]. Viruses also have a significant impact on the growth of grape leaves [5]. Most growers diagnose diseases manually by observing plants with identified diseases only by naked eye. Timely disease identification is very important as a small number of diseased leaves can transmit the disease to other leaves and even to the grapes [6-7]. If growers want a good harvest, then it is very important to detect vine diseases at an early stage to recommend treatment to prevent heavy losses. Although many agricultural experts are skilled in identifying diseases, they also have short comings such as losing focus, resulting in decreased accuracy in disease identification. Therefore, an automated system with precise accuracy is needed. The combination of image processing, pattern recognition, and classification technology can help to overcome or at least reduce the problem of disease identification [8-10]. The aim of this research is to use digital image processing to detect and characterize diseases in grape plants. Identifying diseases in grape plants is a crucial aspect of agriculture as it can aid farmers or agricultural experts in controlling and preventing the spread of diseases. Conventional identification methods such as visual observation can be time-consuming and not always accurate [11-12]. Therefore, image processing techniques have been proposed as an alternative solution for quickly and accurately identifying diseases in grape plants [13-14]. Previous studies have tested the effectiveness of image processing techniques in identifying diseases in grape plants. For example, a study used color segmentation technique to identify leaves of grape plants infected with diseases. The study showed that the color segmentation technique successfully separated infected leaves from healthy ones with a high level of accuracy [15-16]. In addition, research has been conducted using climate-based techniques to identify Downy Mildew and Powdery Mildew. By observing the existing climate conditions, timely diagnosis and accurate detection of plant diseases can be achieved [17]. Other studies have utilized Ghost convolution and Transformer networks to diagnose grape leaves [18], while CNN architecture has been used to identify diseases such as black rot, esca, and isariopsis leaf spot on healthy grape leaves [19]. ResNet-50 (Residual Network-50) is one of the most popular and effective convolutional neural network (CNN) model architectures for image processing tasks, including identifying diseases in grape plants. These blocks allow the network to 'learn' residual information (differences) between the input and output of a layer, enabling deeper and better network training. The main advantage of the ResNet architecture is the use of residual blocks.

Kata Kunci: Grape plant, Grape varieties, Inception ResNet, Leaf detection, ResNet