Evaluation of Robust Covariance Estimation for Object Detection
Fecha
2021-04-07
Autores
Tamayo-Arango, Andres Felipe
Plazas Escudero, David
Vidal-Correa, Juan Pablo
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad EAFIT
Resumen
This work presents an initial approach to the evaluation of robust covariance estimation for object detection (localization) using the “region covariance” technique from the literature. The covariance estimation is performed using the Comedian, Kendall, Spearman and Ledoit and Wolf robust approaches for covariance, and the procedure was also compared using two different matrix norms for estimating dissimilarity. The performance was measured quantitatively using linear regression and Pareto boundaries, yielding the Ledoit and Wolf estimation with best overall performance in object detection in normal and noisy images.