INBaSA: Novel Non-parametric background subtraction algorithm based on and improved pixel based segmentation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

INBaSA: Novel Non-parametric background subtraction algorithm based on and improved pixel based segmentation

Lionel Landry Sop Deffo
  • Fonction : Auteur
  • PersonId : 1143367

Résumé

Applications dealing with images have increased recently. These include video surveillance, person identification, pedestrian detection, tracking, etc. All these operations need to be performed following a certain number of steps of which one of the most important is background subtraction used for motion detection. Despite improvements in several proposed approaches, there is still demand for better performance from users. It is for this reason that, many algorithms have captured our attention among which is the ViBE-based algorithm called EFF-ViBE. While the first (ViBE) suffers from two major drawbacks which are; the apparition of ghost phenomenon as well as poor quality of results in case of complex background. The second (EFF-ViBE) on the other hand while trying to solve those drawbacks introduces a large number of false positives and false negatives and therefore affects the results. This paper presents an approach called INBaSA to address the problems mentioned above. We have introduced new factors such as the mean of temporal gradients, and adaptive radius with offset values, change the pixel counting threshold, and modified the segmentation process.
Fichier principal
Vignette du fichier
INBaSA.pdf (572.5 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03816276 , version 1 (25-10-2022)

Identifiants

  • HAL Id : hal-03816276 , version 1

Citer

Lionel Landry Sop Deffo, Elie Fute Tagne. INBaSA: Novel Non-parametric background subtraction algorithm based on and improved pixel based segmentation. CARI, Oct 2022, Dschang, Cameroon. ⟨hal-03816276⟩

Relations

Collections

AFRIQ
22 Consultations
23 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More