DETECÇÃO DE MICRORGANISMOS AMBIENTAIS EM IMAGENS MICROSCÓPICAS UTILIZANDO REDES NEURAIS CONVOLUCIONAIS

Abstract

The study of environmental microorganisms is essential to characterize the biodiversity found in different places, such as soil and rivers. Isolating microorganisms and identifying their species is a very challenging area, and one that has become increasingly multidisciplinary. A common task is to use characteristics found in the shape of the microorganism (morphology) to identify and classify the species. This task can be done manually or using digital image processing and computer vision techniques. This work proposes the use of a convolutional neural network for object detection with the aim of identifying environmental microorganisms in microscopic images, through the use of a deep machine learning technique and a public database of images of environmental microorganisms. Since the number of images is important to obtain good results in this process, several pre-processings were performed on the images with the aim of increasing the number of images in the original database (data augmentation). With this, the model was trained with both sets, original and augmented, and at the end of the training, validations and tests were carried out with the evaluation metrics, indicating that it is possible to obtain good results in detecting environmental microorganisms with the proposed methodology.