From: Monitoring sustainable development by means of earth observation data and machine learning: a review
SDGs | Field | Main finding | References |
---|---|---|---|
SGD 11 (Sustainable Cities and Communities) | Land use | Tested Sub-PB and Super-PB methodologies to map green spaces. The results showed that Super-PB approach was better for dense urban, sub-urban and rural subsets. However, for lower-resolution images, the Sub-PB approach performed better for dense urban and sub-urban subsets | [136] |
Land change | Developed two CNN approaches: Early Fusion and Siamese Network to detect changes in pairs of images. Overall, the results proved that Siamese Network approach was the most accurate | [144] | |
SGD 12 (Responsible Consumption and Production) | Consumption | Proved that CNNs combined with high-resolution images represent a precise and cost-effective methodology to calculate consumption expenditure and wealth in developing countries | [145] |
SDG 15 (Life on Land) | Land cover | Analysed 15Â years of research on supervised classification methods and found that SVM was the most accurate among NN, RF and Decision Tree | [146] |