Desenvolvimento de uma ferramenta de identificação e estimativa de contagem de moscas-brancas no feijoeiro

Abstract

Based on the principles of One Health, food security requires sustainable practices that reduce pesticide use without compromising production. The common bean (Phaseolus vulgaris), a widely consumed food especially in Latin America and African countries, is affected by viral diseases transmitted by the whitefly Bemisia tabaci (Hemiptera: Aleyrodidae), such as the bean golden mosaic virus, which can cause severe losses. The control of these diseases depends on vector management, which requires more efficient and lower-cost monitoring methods. In this context, this study aimed to develop a computer vision tool capable of assisting in the process of counting whitefly nymphs on common bean leaves, to facilitate monitoring and integrated vector management actions. The methodology involved the development of a mobile application integrated with the YOLO11 computer vision model, trained on an image dataset composed of 1,352 photographs of nymphs on soybean and bean leaves, randomly divided in an 8:1:1 ratio into training, testing, and validation sets. The model was then used to perform detections on bean leaves. The results were organized into three detection scenarios: (i) mixed resolution with macro images, (ii) low resolution with macro images, and (iii) mixed resolution without macro use. For each scenario, considering a confidence threshold of 0.5, the model achieved the following Precision, Recall, and mAP50 values on the validation set: (0.7735, 0.7215, 0.82071), (0.87024, 0.7014, 0.7554), and (0.83371,0.7393, 0.82189), respectively. Based on the results obtained, it can be concluded that the proposed tool presents satisfactory performance for the detection and counting of whitefly nymphs on bean leaves. However, its performance is influenced by image quality, achieving better results with higher resolution images captured in macro mode. Nevertheless, the tool shows potential as an auxiliary instrument for population monitoring of the vector and for supporting further research in the management and control of this insect.