부산대학교 도서관

One of the contributions of computer vision is precision agriculture, which uses high technology to ensure good production and profitability in agribusiness. Its techniques can be applied to identify diseases and pests and diagnose the degree of leaf damage caused by them. In this context, this work seeks to identify the symptoms of mineral deficiencies and pests in coffee crops through computer vision. Therefore, the general objective of this work is to identify mineral deficiencies and pests in coffee plants using computer vision techniques and convolutional neural networks, using a pre-trained AlexNet architecture model available on the Matlab platform. This research is justified by the difficulty, in agriculture, in determining the damage caused by pests and mineral deficiencies accurately through visual analysis, preventing the taking of measures to minimize them. Therefore, the use of Convolutional Neural Networks can present significant results for the accurate classification of diseases in coffee plants, helping to minimize losses in productivity. 117 images were tested with the static classifier, among which the neural network managed to classify 98 correctly. Some images obtained correct classification, but without 100% indication of the real class.