A new algorithm to automatically detect the pith on rough log-end images - INRA - Institut national de la recherche agronomique Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

A new algorithm to automatically detect the pith on rough log-end images

Phuc Ngo
Frederic Mothe
  • Fonction : Auteur
  • PersonId : 1287294
  • IdHAL : fmothe
Fleur Longuetaud

Résumé

X-ray computer tomography has proved to be efficient for measuring internal and external characteristics of logs that are relevant for estimating wood quality at the sawmill. However, this technology remains expensive (two sawmills in France are equipped) and there is a need to provide low-cost tools for smaller sawmills. It could help them become more competitive. There is also a need for new tools available everywhere , for example to estimate wood quality on the harvester, at the road side, or on the log yard (e.g. with mobile-phone cameras). Contrary to X-ray scans, low-cost cameras provide very different images. Moreover few works have been done on such cameras so far. The pith in log end images is an important feature. It is usually required to detect other wood characteristics (for example annual rings) and to process further toward wood quality estimation. The pith's location is a real challenging problem for untreated log ends. In this context, we propose a robust and efficient algorithm to address this challenge. It consists of a mixture between Hough Transform and an approach based on state-of-the-art algorithms, known as Hough-based algorithms. We validated the proposed method on RGB images of Douglas fir taken with a digital camera after harvesting wood in the forest. The obtained results show a better detection of the pith on rough log end images than some state-of-the-art algorithms. The algorithm may process images in real-time which is compatible with sawmill requirements.

Domaines

Informatique
Fichier principal
Vignette du fichier
Article_Freiburg.pdf (6.15 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02275651 , version 1 (20-02-2020)

Identifiants

  • HAL Id : hal-02275651 , version 1

Citer

Rémi Decelle, Phuc Ngo, Isabelle Debled-Rennesson, Frederic Mothe, Fleur Longuetaud. A new algorithm to automatically detect the pith on rough log-end images. 21st International Nondestructive Testing and Evaluation (NDTE) of Wood Symposium, Sep 2019, Freiburg, Germany. ⟨hal-02275651⟩
289 Consultations
294 Téléchargements

Partager

Gmail Facebook X LinkedIn More