Automated logo recognition has emerged as a significant research area in computer vision due to its applications in document authentication, institutional identification, digital archiving, and content-based image retrieval systems. This study proposes a robust hybrid framework for the automated identification of Nigerian Federal University logos by integrating Maximally Stable Extremal Region (MSER) descriptor analysis with a Convolutional Neural Network (CNN)-driven classification model. The research is motivated by the challenges of logo variability, intra-class similarity, illumination changes, and low-resolution image constraints commonly encountered in real-world academic and administrative datasets.

The proposed approach leverages MSER-based region detection for stable feature extraction followed by deep feature learning through CNN architectures to improve classification accuracy and generalization performance. The integration of classical feature-based detection and modern deep learning techniques provides a hybrid solution that enhances robustness under complex visual conditions. Prior research has demonstrated that MSER is highly effective in detecting stable regions in images (Matas et al., 2004), and its adaptation in logo detection pipelines has shown promising results in document analysis systems (Alaei & Delalandre, 2014). Furthermore, CNN-based architectures have consistently achieved superior performance in image recognition tasks due to their hierarchical feature extraction capabilities (Krizhevsky et al., 2017).

Experimental considerations suggest that combining MSER region proposals with CNN classification significantly reduces false positives and improves recognition precision in structured institutional logo datasets. The study contributes to advancing computer vision applications in educational domain recognition systems and provides a scalable framework for national-level institutional logo classification