Aplicación de algoritmo de inteligencia artificial para la detección de cultivos de cacao (Theobroma cacao L.), banano (Musa paradisiaca L.) y palma africana (Elaeis guineensis J.) en la zona norte de las provincias del Guayas y los Ríos
Remote sensing is a crucial tool in land planning, so it is necessary to evaluate the effectiveness of new methods emerging in this field in real environments. In this study, the accuracy of the traditional remote sensing method, Maximum Likelihood (ML), was compared with a method based on artificia...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| التنسيق: | bachelorThesis |
| اللغة: | spa |
| منشور في: |
2023
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | https://repositorio.uteq.edu.ec/handle/43000/7021 |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
إضافة وسم | الملخص: | Remote sensing is a crucial tool in land planning, so it is necessary to evaluate the effectiveness of new methods emerging in this field in real environments. In this study, the accuracy of the traditional remote sensing method, Maximum Likelihood (ML), was compared with a method based on artificial intelligence called Random Forest (RF), in the identification of banana, cocoa and African palm crops, as well as other land covers. Sentinel-2 satellite images obtained through the USGS website were used, which were atmospherically corrected. From these images, NDVI (Normalized Difference Vegetation Index) and RESI (Red-Edge Spectral Indices) spectral indices were calculated and regions of interest (ROIs) were defined on the calculated index layer of each image. ROIs were created from both in situ coordinates and Google Earth. ML classification was performed in the ENVI software environment and RF classification was performed in the QGIS program with Dzetsaka add-on tool. To evaluate the results, traditional confusion matrix, Kappa coefficient and Wilcoxon-Mann-Whitney non-parametric statistical test were used. The results indicated that ML most accurately detected African palm crops, however, the most accurate overall classification (including crops of interest and other cover crops) was obtained with RF. Therefore, it is recommended to continue investigating the potential of RF under more favorable conditions, such as images recorded during dry seasons and with a larger number of available samples. Keywords: Maximum Likelihood, Random Forest, Crop, NDVI, RESI. |
|---|
Algoritmo de inteligencia artificial para la detección de cultivos de cacao (teobroma cacao l.), banano (musa paradisiaca l.) y palma africana (elaeis guineensis j.).