Detección de individuos y grupos de palmas de cera (Ceroxylon sp) en imágenes satelitales de alta resolución, mediante herramientas de Aprendizaje Profundo en ArcGIS Pro

It was carried out the training of a model based on Deep Learning and RetinaNettype convolutional neural networks, for detection of individuals and groups of wax palms (Ceroxylon sp) in high-resolution satellite images, using the tools available for object detection in ArcGIS Pro. The model was gene...

Full description

Saved in:

Bibliographic Details
Main Authors:	Vecino Salcedo, Cristian Fernando, Ramos Patiño, Juan Pablo
Other Authors:	Carvajal Vanegas, Andrés Felipe
Format:	Trabajo de grado (Pregrado y/o Especialización)
Language:	spa
Published:	Universidad Antonio Nariño 2022
Subjects:	Palma de cera Aprendizaje profundo Redes neuronales convolucionales Detección de objetos Sensores remotos. Wax palm Deep learning Convolutional neural networks Object detection Remote sensing
Online Access:	http://repositorio.uan.edu.co/handle/123456789/7129
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	It was carried out the training of a model based on Deep Learning and RetinaNettype convolutional neural networks, for detection of individuals and groups of wax palms (Ceroxylon sp) in high-resolution satellite images, using the tools available for object detection in ArcGIS Pro. The model was generated from a first training phase with sampling accomplished on a sector of isolated palms and visually identified palm groves in the zone of the Cocora valley in Salento, department of Quindío; subsequently, and then carry out the model validation in the entire zone and optimizing the training and detection parameters, automatic identification of wax palms was performed in the implementation zone, corresponding to Alto de Toche and La Ceja jurisdiction, municipalities of Ibagué and Cajamarca, department of Tolima; obtaining an average modelling precision score of 0.74, and a percentage of less than 2% of omitted individuals and false detections in pasture areas, and greater than 30% in areas of forest cover.