Citas bibligráficas
Blanco, A., Sanchez, J. (2024). Desarrollo de un equipo electrónico portátil orientado a la identificación y clasificación de especies de árboles maderables basado en el procesamiento digital de imágenes de hojas e Inteligencia Artificial [Tesis, Universidad Peruana de Ciencias Aplicadas (UPC)]. http://hdl.handle.net/10757/673263
Blanco, A., Sanchez, J. Desarrollo de un equipo electrónico portátil orientado a la identificación y clasificación de especies de árboles maderables basado en el procesamiento digital de imágenes de hojas e Inteligencia Artificial [Tesis]. PE: Universidad Peruana de Ciencias Aplicadas (UPC); 2024. http://hdl.handle.net/10757/673263
@misc{renati/1297873,
title = "Desarrollo de un equipo electrónico portátil orientado a la identificación y clasificación de especies de árboles maderables basado en el procesamiento digital de imágenes de hojas e Inteligencia Artificial",
author = "Sanchez Torres, Jose Alejandro",
publisher = "Universidad Peruana de Ciencias Aplicadas (UPC)",
year = "2024"
}
This dissertation proposes the development of a portable equipment for the classification of 11 species of trees present in the Peruvian jungle using tree’s leaves as input. Currently, the identification method is carried out by a botanist, who in the first instance determines the species by visual inspection and, if necessary, analyzes a sample in a laboratory with a dendrological manual. For this reason, the process is inefficient in time, cost, and precision. The state-of-the-art solutions present the following difficulties: no detection of Peruvian timber species, but instead use public datasets or build their own datasets with fruit, medicinal or non-Peruvian plants; some works have high error rates; and most proposals use mobile applications instead of their own portable devices for in situ image acquisition. In that regard, the algorithm that integrates the proposed equipment was developed with a database that contains 11 Peruvian timber species and in conjunction with the equipment designed considering ideal capture conditions and controlled lighting, reduces the error percentage to 1.36%. This algorithm acquires the image of a leaf entered into the equipment in its first stage. Then, it performs the segmentation of the captured image using color filters, binarization thresholds and morphological operations to finally predict the tree specie to which it belongs using an already established CNN model. Said software is complemented by hardware that is made up of a raspberry pi 4 with an interactive graphical interface for the prediction sample. An accuracy of 98.64% was obtained using the MobileNet model and a Kappa cohen index of 0.9733, which indicates a very high level of agreement and shows that the equipment’s results match those issued by the specialist.
Este ítem está sujeto a una licencia Creative Commons Licencia Creative Commons