Comparing the Segment Anything Model with Region Growing Algorithms in the detection of irrigated croplands

The advance of remote sensing and geotechnologies has helped to solve agricultural-related problems, especially those connected to management practices such as irrigation. Image segmentation techniques, for example, bring the possibility of identifying areas and borders of irrigated croplands,a fact...

Descripción completa

Detalles Bibliográficos
Autores: Petrone, Felipe Gomes, Da Silva, Darlan Teles, Maia, Aluizio Brito, Sanches, Ieda Del'Arco, Dantas Chaves, Michel Eustáquio [UNESP], Garcia Fonseca, Leila Maria, Körting, Thales Sehn, Adami, Marcos
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2024
País:Brasil
Institución:Universidade Estadual Paulista (UNESP)
Repositorio:Repositório Institucional da UNESP
Idioma:inglés
OAI Identifier:oai:repositorio.unesp.br:11449/306011
Acceso en línea:http://dx.doi.org/10.14393/rbcv76n0a-72592
https://hdl.handle.net/11449/306011
Access Level:acceso abierto
Palabra clave:Image Segmentation
Irrigated Croplands
Remote Sensing Images
Descripción
Sumario:The advance of remote sensing and geotechnologies has helped to solve agricultural-related problems, especially those connected to management practices such as irrigation. Image segmentation techniques, for example, bring the possibility of identifying areas and borders of irrigated croplands,a factor that can enhance monitoring and yield estimates. In this research field, a recent innovation is the Segment Anything Model (SAM) algorithm. Thus, this study aimed to compare SAM with two well-known remote sensing image segmentation algorithms, Region Growing and Baatz-Schape, in order to delineate irrigated agricultural lands in the Brazilian semiarid region. The findings indicate that SAM has the potential to generate homogeneous segments when examining such lands. However, it requires refinements in order to distinguish fields with varying crops and to improve the high computational cost of SAM, especially for big data. Additionally, the choice and testing of parameters are crucial for the optimal performance of segmentation algorithms.