|
Título:
|
Computer vision techniques for greenness identification and obstacle detection in maize fields
|
|
Autores:
|
Campos Silvestre, Yerania
|
|
Tipo de documento:
|
texto impreso
|
|
Editorial:
|
Universidad Complutense de Madrid, 2017-06-21
|
|
Dimensiones:
|
application/pdf
|
|
Nota general:
|
info:eu-repo/semantics/openAccess
|
|
Idiomas:
|
|
|
Palabras clave:
|
Estado = No publicado
,
Materia = Ciencias: Informática: Inteligencia artificial
,
Tipo = Tesis
|
|
Resumen:
|
There is an increasing demand in the use of Computer Vision techniques in Precision Agriculture (PA) based on images captured with cameras on-board autonomous vehicles. Two techniques have been developed in this research. The rst for greenness identi cation and the second for obstacle detection in maize elds, including people and animals, for tractors in the RHEA (robot eets for highly e ective and forestry management) project, equipped with monocular cameras on-board the tractors. For vegetation identi cation in agricultural images the combination of colour vegetation indices (CVIs) with thresholding techniques is the usual strategy where the remaining elements on the image are also extracted. The main goal of this research line is the development of an alternative strategy for vegetation detection. To achieve our goal, we propose a methodology based on two well-known techniques in computer vision: Bag of Words representation (BoW) and Support Vector Machines (SVM). Then, each image is partitioned into several Regions Of Interest (ROIs). Afterwards, a feature descriptor is obtained for each ROI, then the descriptor is evaluated with a classi er model (previously trained to discriminate between vegetation and background) to determine whether or not the ROI is vegetation...
|
|
En línea:
|
https://eprints.ucm.es/id/eprint/46034/1/T39509.pdf
|
