Desarrollo de una aplicación móvil para la detección temprana de plagas en cultivos de maíz (zea mays l.) Empleando técnicas de visión artificial basada en aprendizaje profundo.

Early detection of pests in maize (Zea mays L.) crops represents a critical challenge for agriculture, directly affecting productivity, crop quality, and food security. With technological advancements, artificial vision and deep learning techniques emerge as promising solutions to this issue. For this reason, the "PlagaScan" mobile application was developed in this project, enabling pest detection in maize crops and providing diagnostic support for inexperienced farmers struggling to identify pests in their crops. The mobile application offers three main functionalities: real-time pest diagnosis, diagnosis through images uploaded from the gallery, and pest control by providing relevant information on possible control or treatment methods for the identified pest. Diagnoses are generated by a deep learning computational model based on the MobileNet-v2 convolutional neural network architecture, which has been trained and validated with a total of 750 images previously prepared and preprocessed. This model is designed to detect five types of pests: white grubs, cutworms, wireworms, mole crickets, and lepidopteran worms. During the model's training, adjustments were made to the hyperparameters, including a batch size of 16 and training for 150 epochs. The Mobile-D methodology, specifically designed for mobile application development, was chosen for the application's development. The application provides a very high performance, with 56.42% accuracy in diagnoses, making it a useful tool for farmers with little experience in detecting pests in maize.

• Autor principal: Zambrano Monserrate , Dennisse Paola (author)
• Format: bachelorThesis
• Idioma: spa
• Publicat: 2024
• Matèries: Aprendizaje profundo; Detección de objetos; Plagas de cultivo de maíz
• Accés en línia: https://repositorio.uteq.edu.ec/handle/43000/7224