The tomato (Solanum lycopersicum (Solanaceae)) is particularly susceptible to Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae), a pest that directly and profoundly influences tomato yields. Consequently, the early detection of T. absoluta damage intensity on leaves using machine learning or artificial intelligence-based algorithms is crucial for effective pest control. In this ground-breaking study, the galleries generated by T. absoluta were examined via field images using the Decision Trees (DTs) algorithm, a machine learning method. The unique advantage of DTs over other algorithms is their inherent capacity to identify complex and vague shapes without the necessity of feature extraction, providing a more streamlined and effective approach. The DTs algorithm was meticulously trained using pixel values from the leaf images, leading to the classification of pixels within regions with and without galleries on the leaves. Accordingly, the gallery intensity was determined to be 9.09% and 35.77% in the test pictures. The performance of the DTs algorithm, as evidenced by a high precision and an accuracy rate of 0.98 and 0.99 respectively, testifies to its robust predictive and classification abilities. This pioneering study has far-reaching implications for the future of precision agriculture, potentially informing the development of advanced algorithms that can be integrated into autonomous vehicles. The integration of DTs in such applications, due to their unique ability to handle complex and indistinct shapes without the need for feature extraction, sets the stage for a new era of efficient and effective pest control strategies.
8 December 2023 in Bursa Uludağ Üniversitesi Ziraat Fakültesi Dergisi
Sunflower powdery mildew (Golovinomyces cichoracearum (DC.) V.P. Heluta) is a substantial threat to sunflower crops, causing significant yield loss. Traditional identification methods, based on human observation, fall short in providing early disease detection and quick control. This study presents a novel approach to this problem, utilizing machine learning for the early detection of powdery mildew in sunflowers. The disease severity levels were determined by training a Decision Trees model using matrix of soil, powdery mildew, stems, and leaf images obtained from original field images. It was detected disease severity levels of 18.14% and 5.56% in test images labeled as A and C, respectively. The model's demonstrated accuracy of 85% suggests high proficiency, indicating that machine learning, specifically the DTs model, holds promising prospects for revolutionizing disease control and diseases prevention in agriculture.
24 July 2024 in Türk Tarım ve Doğa Bilimleri Dergisi
Bu çalışma, hünnap bitkilerinde gözlemlenen Metcalfa pruinosa zararlısının erginlerini tespit etmek amacıyla YOLOv5 algoritmasının v5s, v5m ve v5l modellerini kullanmayı hedeflemiştir. Böylelikle, tarımsal mücadelenin başlama anını belirlemek için kullanılan cihazlar ve zararlı popülasyon yoğunluğuna göre ilaçlama yapan robotik sistemler için bir kaynak teşkil etmektedir. Modellerin eğitimi için kullanılacak görüntüler elde edildikten sonra, veri artırımı yöntemleri kullanılarak veri setleri genişletilmiş ve görüntüler Roboflow kullanılarak etiketlenmiştir. Ardından, bu veriler kullanılarak modeller eğitilmiş ve eğitilen modellerin box_loss, obj_loss, precision, recall, mAP_0.5 ve mAP_0.5:0.95 gibi performans metrikleri analiz edilmiştir. YOLOv5s modelinde, box_loss ve obj_loss performans metriklerinin sırasıyla 0.02858 ve 0.0055256 değerleri ile en yüksek olduğu bulunmuştur. YOLOv5m modelinde, recall performans metriğinin 0.98127 değeri ile en yüksek olduğu tespit edilmiştir. YOLOv5l modelinde ise precision, mAP_0.5 ve mAP_0.5:0.95 performans metriklerinin sırasıyla 0.98122, 0.99500 ve 0.67864 değerleri ile en yüksek olduğu belirlenmiştir. Sonuç olarak, YOLOv5l modeli diğerlerine göre daha yüksek doğruluk sergilemektedir. YOLOv5l modelinin, Metcalfa pruinosa zararlısının tespiti için yeterli olduğu düşünülmektedir.