Segmentação automática de objetos em imagens resultantes de experimentos de memória utilizando técnicas de processamento digital de imagens

Abstract

The object recognition task uses models based on their natural curiosity to explore new items, which makes it a widely used tool in research to investigate different stages of memory through behavior. This paradigm ranges from acquisition to recall and reconsolidation, providing a diverse framework for exploring multiple aspects of memory. However, manual analysis of results can be time consuming and susceptible to potential bias. In this context, this work aims to develop a system that automates the identification of coordinates of objects present in recognition tasks with animal models, more specifically, with rodents. This information is essential for the application of software that monitors the voice and the exploration time of each item by the animal, allowing a more agile and objective analysis of the tests. With this, investigators will be able to direct their focus to other crucial stages of the investigation. The results obtained after applying image processing techniques provided an accuracy of 99.01% in identifying items, showing an average difference of only 4.07 pixels compared to manual marking of the center of the object.