Au cours de ce travail nous avons utilisé les potentialités de différentes modalités d'imagerie, que nous appliquons au domaine du végétal afin de contribuer au phénotypage haut-débit des semences et plantules. Nous nous somme principalement consacré à la recherche de réponses à deux problématiques spécifiques et importantes de ce secteur. Nous commençons par montrer l'applicabilité de l'imagerie en lumière visible inactinique et de l'imagerie thermographique passive pour imager le développement des semences et plantules, un phénomène biologique se déroulant normalement dans le sol et l’obscurité. Nous présentons nos apports à ce type d’imagerie, au travers de nos contributions à la conception et à la réalisation d’un système de vision en imagerie visible inactinique, ayant pour finalité la réalisation de mesures individualisées automatisées sur les semences, les plantules et les organes des plantules. Ce système gère les croisements de plantules, via l'utilisation originale de la diffusion anisotrope, ce qui nous a permis de multiplier, sans perte d'information, les débits par dix. De plus, ce système réalise la séparation des organes au moyen d’un critère générique basé sur le gravitropisme. La validation des algorithmes de traitement d'images du système de vision utilise des voies originales (simulation numérique et test de l'influence de l'incertitude via simulation agronomique). L’imagerie thermographique, qui capte le rayonnement thermique passif des objets, nous permet de visualiser et de mesurer les semences et plantules dans l'obscurité. Elle permet aussi de réaliser la segmentation et le suivi des organes de plantules. Cette technologie d'imagerie nous a aussi permis de montrer la faisabilité d'un dosage non destructif de teneur en sucre des organes de plantules de betterave. Ensuite nous proposons une méthodologie générique permettant la conception de capteurs bas-coût spectralement optimisés, en fonction de tâches applicatives déterminées. Cette méthodologie utilise la théorie de l’information, pour extraire de l’imagerie hyperspectrale, relativement coûteuse, l’information utile à la conception des capteurs dédiés bas-coût. L’intérêt de cette méthodologie pour le phénotypage des plantes est montré et justifie le transfert de celle-ci au monde de la recherche en biologie végétale. / Along this work, we have used the potentiality of different modalities of imagery that we apply to the plant domain so as to contribute to the high-throughput phenotyping of seeds and seedlings. We have mainly committed ourselves to the search for answers to two specific and important problematic in this domain. We begin by showing the applicability of visible imaging using an inactinic light and passive thermographic imaging to image the development of seeds and seedlings, a biological phenomenon usually occurring in soil and darkness. We present our contributions to this type of imaging through our contributions to the conception and the realization of a vision system using visible inactinic imaging, whose finality is the realization of individualized automated measurement on the seeds, the seedlings and the organs of the seedlings. This system handle seedling crossing, through the original use of anisotropic diffusion, which allowed us to multiply, without information loss, the output by ten. Furthermore, this system carries out the separation of the organs by means of a generic criterion based on gravitropism. The validation of the image processing algorithms of the vision system use original ways (numerical simulation and test of the influence of the uncertainty through agronomic simulation). Thermographic imaging, which captures the passive heat radiation of objects, allows us to visualize and to measure seeds and seedlings in the darkness. It also allows realizing the segmentation and the tracking of the organs of seedlings. This imaging technology also allowed us to demonstrate the feasibility of a non-destructive determination of sugar quantity in organs of beet seedlings. We then propose a generic methodology that allows the conception of spectrally optimized low-cost sensors, according to determined application tasks. This methodology uses information theory, to extract from, relatively expensive, hyperspectral imaging, the information needed for the conception of the dedicated low-cost sensors. The interest of this methodology for plant phenotyping has been shown and justifies its transfer to the world of research in plant biology.

<p>Izvod<br />IZ<br />Pet hibrida suncokreta (Sremac, Oliva, Cepko, NS-H-111 i Sumo 2 OR) gajena su tokom<br />vegetacionih sezona 2010. i 2011. godine na oglednom polju Instituta za ratarstvo i<br />povrtarstvo, Novi Sad, na Rimskim &scaron;ančevima i na oglednom polju Poljoprivredne stručne<br />službe Zrenjanin. Laboratorijski deo ogleda izveden je u Laboratoriji za ispitivanje semena<br />Instituta za ratarstvo i povrtarstvo iz Novog Sada i Biohemijskoj laboratoriji<br />Poljoprivrednog fakulteta u Novom Sadu. Cilj istraživanja je bio da se utvrdi da li semenski<br />materijal, poreklom sa istog lokaliteta i iste parcele, u procesu dorade semena tj. odvajanja<br />semena po frakcijama menja kvalitet, kao i da li ovakav način dorade semenskog materijala<br />utiče na prinos i komponente prinosa semena, kao i na sadržaj i prinos ulja odabranih<br />hibrida suncokreta. Takođe, cilj je bio i da se utvrdi da li nedostatak primarnog korena kod<br />ponika suncokreta dovodi do statistički značajnog smanjenja prinosa semena i ulja, kao i<br />kakav uticaj ima na komponente prinosa suncokreta. Analizirane su sledeće osobine:<br />klijavost i vigor semena, enzimska aktivnost, nicanje u polju, nakupljanje suve materije (po<br />biljnim organima i ukupno po biljci), visina stabla, prečnik glave, prinos semena, masa 1000<br />semena, sadržaj ulja i prinos ulja. Statistička obrada podataka izvr&scaron;ena je primenom analize<br />varijanse trofaktorijalnog ogleda, prema modelu podeljenih parcela. Iz tabele analize<br />varijanse je prikazana verovatnoća značajnosti razlika po F-testu, a na osnovu uče&scaron;ća u<br />sumi kvadrata tretmana određen je procentualni udeo svakog faktora u ukupnoj<br />varijabilnosti. Takođe su izračunate i LSD vrednosti za poređenje razlika između tretmana<br />posmatranog faktora, na pragovima značajnosti od 1 i 5%. Utvrđena je i korelaciona<br />zavisnost posmatranih osobina. Rezultati analize varijanse pokazali su visoko značajno<br />uče&scaron;će glavnih faktora (godine, lokaliteta, frakcije semena, tipa ponika), kao i njihovih<br />međusobnih interakcija za većinu ispitivanih osobina. Na vrednosti ispitivanih osobina<br />najveći uticaj imala je godina istraživanja, a zatim lokalitet. Frakcija semena je imala visoko<br />značajan uticaj na klijavost semena, nicanje u polju, nakupljanje suve materije, prinos<br />semena i prinos ulja. Tip ponika je imao visoko značajan uticaj na enzimsku aktivnost</p> / <p>Five sunflower hybrids (Sremac, Oliva, Cepko, NS-H-111, and Sumo 2 OR) were grown in<br />2010 and 2011, on experimental fields of the Institute of Field and Vegetable Crops Novi<br />Sad - Rimski &Scaron;ančevi, and experimental fields of the Agricultural Service Zrenjanin.<br />Laboratory tests were conducted at Laboratory for Seed Testing of the Institute of Field<br />and Vegetable Crops Novi Sad, and Biochemical Laboratory of the Faculty of Agriculture,<br />University of Novi Sad. The aim of this research was to determine whether seed material<br />from the same locality and the same field lot changes quality during seed processing, i.e.<br />dividing seed lot into fractions, and additionally whether this type of seed processing<br />affects seed yield, seed yield components, oil yield, and oil content of the selected<br />sunflower hybrids. The aim was also to determine if missing primary root in sunflower<br />seedlings results in statistically significant decrease of seed and oil yield, and establish how<br />it influences sunflower yield components. The following traits were examined: seed<br />germination and vigour, enzymatic activity, field emergence, dry matter accumulation (in<br />plant organs and in the whole plant), stem height, head diameter, seed yield, 1000 seed<br />mass, oil content and oil yield. Statistical analysis of data was performed by analysis of<br />variance of the trifactorial trial using the split-plot design model. Table of analysis of<br />variance shows the probability of significance of differences by F-test, and based on the<br />participation in the treatment sum of squares, percentage ratio of each factor was<br />calculated in the total variability. LSD values at 1 and 5% were computed to compare<br />differences between treatments of the observed factor. Correlation dependence between<br />the observed traits was determined. Results of the analysis of variance showed a highly<br />significant participation of the main factors (year, locality, seed fraction, type of seedling),<br />and their mutual interactions for the majority of the examined traits. Year and locality had<br />the highest effect on the value of the examined traits. Seed fraction had highly significant<br />effect on seed germination, emergence, dry matter accumulation, seed yield, and oil yield.<br />Type of seedling had a highly significant effect on the enzymatic activity, seed yield, oil<br />content, and oil yield.</p>