RT Dissertation/Thesis
T1 Contributions to the segmentation of moving objects in video sequences
T2 Contribuciones a la segmentación de objetos en movimiento en secuencias de vídeo
A1 De Castro Lopes Martins Pinto Ferreira, María Isabel
K1 33 Ciencias Tecnológicas
K1 2290 Física Altas Energías
K1 3304.99 Otras
AB Developing robust and universal methods for unsupervised segmentation of moving objects in video sequences has proved to be a hard and challenging task. State-of-the-art methods show good performance in a wide range of situations but none has been able to fully deal with complex and challenging scenarios that include poor lighting conditions, sudden illumination changes, shadows and parasitic background motion. There are several approaches for segmenting foreground from background. Recent research has shown that methods appear to be complementary in nature, with the best-performing methods being beaten by combining several of them. In this PhD work we want to further explore some of the most efficient and widely used approaches (such as GMM), alone or in combination with new techniques, to propose more robust algorithms.The experiments for this research will be conducted on the complete set of videos provided in the CDnet 2014 Dataset, consisting of 53 videos depicting indoor and outdoor scenes captured in different scenarios and with different cameras, containing a wide range of different challenges. Testing and evaluation will be performed using the ground truth segmentation provided along with the videos. This approach aims at producing repeatable experiments that may contribute to the consistent development of the state-of-the-art in this domain.
YR 2020
FD 2020-02-03
LK http://hdl.handle.net/11093/1427
UL http://hdl.handle.net/11093/1427
LA eng
DS Investigo
RD 26-sep-2023