Accéder directement au contenu Accéder directement à la navigation
Chapitre d'ouvrage

Anatomofunctional bimodality imaging for plant phenotyping: An insight through depth imaging coupled to thermal imaging

Abstract : Figure 9.2 The stereovision system and its possible occlusions. (a) Principle scheme of a stereovision system with two cameras. (b) Illustration of occlusions. Nayar and Nakagawa, 1994). At each increment of the translation stage, the vision sensor acquires an image that is blurred for points out of the focused plane and sharp for points in the focus plane. In each point of the object, blur appearance is like the application of a low-pass filter on the focused image. The quantification of focus on each point of the object in an image can be done by quantifying the amount of high spatial frequencies (Xiong and Shafer, 1993; Nayar and Nakagawa, 1994; Subbarao and Choi, 1995; Martinez Baena et al., 1997; Choi and Yun, 2000; Helmli and Scherer, 2001; Ahmad and Choi, 2005, 2007; Malik and Choi, 2007, 2008; Minhas et al., 2009). All these focus measures work on highly textured objects. In each object point, the evolution of focus is computed as a function of the position of the translation stage. The position of the translational stage corresponding to the maxima of the focus measure gives the depth of the considered point. In depth from focus methods, the vision sensor can be a low-cost webcam. Due to the translation stage, depth from focus systems are cumbersome and hardly usable in the field. In depth from defocus methods the lens parameters (aperture, focal) are changed so that the scene does not have to be translated. The defocus measure consists in approximating the PSF in each point of the scene. The PSF can be approximated only for textured subwindows and for objects out of field depth. For each scene point, the PSF approximation is done locally in subwindows with a statistical framework (Rajagopalan and Chaudhuri, 1999; Schechner and Kiryati, 1999; Farid and Simoncelli, 1998) or with deterministic optimization (Xiong and Shafer, 1995; Gokstorp, 1994; Favaro and Soatto, 2000; Favaro et al., 2003; Trouvé et al., 2011). Depth from defocus methods must use vision sensors with well-known parameters (aperture, focal). The use of low-cost webcams for these methods is therefore prohibited. The depth from defocus systems are suitable for use in greenhouse on highly textured and nonmovable plants.
Liste complète des métadonnées

Littérature citée [130 références]  Voir  Masquer  Télécharger
Contributeur : Marie-Françoise Gerard <>
Soumis le : jeudi 9 avril 2020 - 15:04:45
Dernière modification le : mercredi 8 juillet 2020 - 12:43:01


  • HAL Id : hal-02536794, version 1


Yann Chéné, Etienne Belin, François Chapeau-Blondeau, Valerie Caffier, Tristan Boureau, et al.. Anatomofunctional bimodality imaging for plant phenotyping: An insight through depth imaging coupled to thermal imaging. S Dutta Gupta; Yasuomi Ibaraki. Plant Image Analysis Fundamentals and Applications, CRC Press, 2014, 9780429072345. ⟨hal-02536794⟩



Consultations de la notice