Mining activities are the leading cause of deforestation, land-use changes, and pollution. Land use/cover mapping in Vietnam every five years is not useful to monitor land covers in mining areas, especially in the Central Highland region. It is necessary to equip managers with a better tool to monitor and map land cover using high-resolution images. Therefore, the authors proposed using the U-Net convolutional network for land-cover classification based on multispectral Unmanned aerial vehicle (UAV) image in a mining area of Daknong province, Vietnam. An area of 0.5kmx0.8km was used for training and testing seven U-Net models using seven optimizer function types. The final U-Net model can interpret six land cover types: (1) open-case mining lands, (2) old permanent croplands, (3) young permanent croplands, (4) grasslands, (5) bare soils, (6) water bodies. As a result, two models using Nadam and Adadelta optimizer function can be used to classify six land cover types with accuracy higher than 83%, especially in open-case mining lands and polluted streams flowed out from the mining areas. The trained U-Net models can potentially update new land cover types in other mining areas towards monitoring land cover changes in real-time in the future.

Convolutional neural network (CNN) is a widely used method in solving classification and regression applications in industries, engineering, and science. This study investigates the optimizing capability of a swarm intelligence algorithm named moth flame optimizer (MFO) for the optimal search of a CNN hyper-parameters (values of filters) and weights of fully connected layers. The proposed model was run with a 3-dimensional dataset (7 width x 7 height x 12 depth), which was constructed through including seven neighbor pixels (vertically and horizontally) from landslide location and 12 predictor variables. Muong Te district, Lai Chau province, Viet Nam was selected as the case study, as it had recently undergone severe impacts of landslides and flash floods. The performance of this proposed model was compared with conventional classifiers, i.e., Random forest, Random subspace, and CNN-optimized Adaptive gradient descend, by using standard metrics. The results showed that the CNN-optimized MFO (Root mean square error = 0.3685, Mean absolute error = 0.2888, Area under Receiver characteristic curve = 0.889 and Overall accuracy = 80.1056%) outperformed the benchmarked methods in all comparing indicators. Besides, the statistical test of difference was also carried out by using the Wilcoxon signed ranked test for non-parametric variables. With these statistical measurements, the proposed model could be used as an alternative solution for landslide susceptibility mapping to support local disaster preparedness plans.

In regular convolutional neural networks (CNN), fully-connected layers act as classifiers to estimate the probabilities for each instance in classification tasks. The accuracy of CNNs can be improved by replacing fully connected layers with gradient boosting algorithms. In this regard, this study investigates three robust classifiers, namely XGBoost, LightGBM, and Catboost, in combination with a CNN for a land cover study in Hanoi, Vietnam. The experiments were implemented using SPOT7 imagery through (1) image segmentation and extraction of features, including spectral information and spatial metrics, (2) normalization of attribute values and generation of graphs, and (3) using graphs as the input dataset to the investigated models for classifying six land cover classes, namely House, Bare land, Vegetation, Water, Impervious Surface, and Shadow. The results show that CNN-based XGBoost (Overall accuracy = 0.8905), LightGBM (0.8956), and CatBoost (0.8956) outperform the other methods used for comparison. It can be seen that the combination of object-based image analysis and CNN-based gradient boosting algorithms significantly improves classification accuracies and can be considered as alternative methods for land cover analysis.

Adaptive Neuro-Fuzzy Inference System (ANFIS) is a robust method in solving non-linear classification by employing a human-readable interpretation manner. This paper verified a hybrid model, named WANFIS, where Whale Optimization Algorithm (WOA) was used for feature selection and tuning parameters of the ANFIS for land-cover classification. Hanoi, the capital of Vietnam, was selected as a case study, because of its complex surface morphology. The model was trained and validated with different data sets, which were subsets of the segmented objects from SPOT 7 satellite data (1.5 m in panchromatic and 6 m multiple spectral bands). The image segmentation was carried out by using PCI Geomatics software (evaluation version), and output objects with associated spectral, shape, and metric information were selected as input data to train and validate the proposed model. For accuracy assessment, the performance of the model was compared to several benchmarked classifiers by using standard statistical indicators such as Receiver Operator Characteristics, Area under ROC, Root Mean Square Error, Absolute Mean Error, Kappa index, and Overall accuracy. The results showed that WANFIS outperformed the other in almost all training data sets for both operations. It could be concluded that the examination of the classification model in different training data sizes is significant, and the proper determination of predictor variables and training sizes would improve the quality of classification of remotely sensed data.

This study aims at investigating the balance between exploration and exploitation search capability of a newly developed Salp swarm optimization algorithm (SSA) for fine-tuning parameters of a three-hidden-layer neural network. The landslide study was selected as a thematic application, and a mountainous area of Vietnam was chosen as a case study. A training dataset with thirteen predictor variables and historical landslide occurrences from the study area were used to train and validate the model. The experiments showed an improvement in several statistic measurements such as Root mean square error = 0.3732, Overall accuracy = 79.35%, Mean absolute error = 0.3075, and Area under Receiver operating characteristic = 0.886 in comparison to conventional benchmark methods. Based on the results, the use of SSA would enhance the search efficiency and could be used as an alternative optimizer for a multiple hidden layer neural network for landslide application as well as for other natural hazard analysis.