Detecção e identificação de doenças em folhas utilizando redes neurais

Abstract

In this study, an approach using neural networks for the detection and classification of diseases in soybean leaves based on images was explored. The main objective of the research was to develop a model capable of analyzing soybean leaf images, identifying different disease classes, and determining the overall health of the plants. The significance of this work lies in its application to precision agriculture, aiming for more effective monitoring and care of crops. The methodology employed included the extraction of features from leaf textures, using techniques such as Histogram of Oriented Gradients (HOG) for identifying textural patterns and edges, as well as obtaining color histograms in the HSV (Hue, Saturation, and Brightness) and RGB (Red, Green, and Blue) components of the images.The results obtained demonstrated promising performance of the proposed model in classifying different disease classes and determining the health of soybean plants. An accuracy rate of 87.54% in classification was achieved, indicating a solid rate of correct classification, even in the face of the complexity of the classes and variability in the images. In summary, this research offers valuable insights for disease detection in soybean plants through neural networks. The results underscore the feasibility and relevance of this approach for precision agriculture and crop health monitoring.The analysis of the confusion matrix reveals valuable information about areas that can be improved, especially in classes with lower representation, emphasizing the importance of future optimizations. This encompasses increasing data collection and expanding the training image dataset, as well as making more detailed adjustments to the hyperparameters of the neural networks. In the future, the exploration of convolutional neural networks and the enrichment of the database have the potential to further contribute to the accuracy and effectiveness of the model, driving significant advancements in disease detection and classification in plants.