DSpace Coleção:https://repositorio.uema.br/jspui/handle/123456789/22012026-09-21T22:55:36Z2026-09-21T22:55:36ZAutomatização do reconhecimento de células germinativas em imagens histológicas de ovários de peixes: estudo de caso de peixes encontrados no estado do Maranhãohttps://repositorio.uema.br/jspui/handle/123456789/62502026-09-04T19:33:52Z2025-02-17T00:00:00ZTítulo: Automatização do reconhecimento de células germinativas em imagens histológicas de ovários de peixes: estudo de caso de peixes encontrados no estado do Maranhão
Abstact: The state of Maranhão has significant fishing activity, making it essential to monitor the
reproductive biology of fish for sustainable management. This work presents an automated
approach for recognizing germ cells in histological images of fish ovaries, using digital image
processing and supervised learning. The objective is to develop a tool that aids in the efficient
analysis of gonads, contributing to knowledge in the field and better management of fishing
resources. The methodology uses the Canny edge detection algorithm to segment cells and
supervised learning techniques to classify them, overcoming the limitations of traditional
manual methods. The results show gains in batch processing speed (50 images/minute) and
precision (56% accuracy) with STERapp. The solution aims to increase the precision and
reproducibility of analyses, positively impacting fishing sustainability in Maranhão and similar
regions2025-02-17T00:00:00ZSistema Web para delegacias com geração de imagens fotorrealistas baseadas em Inteligência Artificial a partir de relatos textuaishttps://repositorio.uema.br/jspui/handle/123456789/62492026-09-04T19:41:58Z2026-01-01T00:00:00ZTítulo: Sistema Web para delegacias com geração de imagens fotorrealistas baseadas em Inteligência Artificial a partir de relatos textuais
Abstact: This work presents the development of a web application designed for police station
management, focusing on integrating artificial intelligence (AI) to generate photorealistic
images from textual descriptions to enhance criminal investigations. Built on a client-server
architecture, the system employs Next.js for the front-end, Flask for the back-end, and the
Stable Diffusion model for image generation. The methodology combined applied, exploratory,
and experimentais approaches with incremental prototyping to ensure scalability and usability.
Requirements analysis identified gaps in virtual police station systems, such as the lack of
advanced visual tools, guiding the development of features like creation, management, and
finalization of incident reports, secure JWT authentication, and AI integration. System
modeling utilized UML, including use case, class, sequence, activity, and deployment
diagrams. Integration and functional tests, conducted with tools like Postman and Cypress,
validated the application's efficiency and robustness. The results demonstrate that the system
meets its objectives, providing an innovative solution that streamlines suspect identification and
improves incident reporting efficiency, with potential for future integration with biometric
databases and further AI model adaptation2026-01-01T00:00:00Z