Pai-Hui Hsu

Human-driven changes in land use and global climate are profoundly reshaping our planet; combined regional effect on hazards remains poorly understood. This study examines the connection between changes in land cover, shifts in Koppen-Geiger (KG) climate zones, and disaster occurrences in Taiwan from 2001 to 2020. We combined satellite data of land cover, surface temperature, and precipitation with verified disaster records. KG climate zones were mapped each year by applying standard classifications to temperature and precipitation data at 1 km spatial resolution. We tracked changes over time and analyzed where they overlapped with disaster events. Our results show an increase in woody savannas and a decline in wetlands, which coincided with a marked expansion of tropical climates and a contraction of temperate climates. Notably, areas where the climate zones shifted-covering 15% of the land- experienced 2.8 times more disasters. localized warming of over 1 °C was strongly linked to forest loss, while extreme rainfall events were correlated with expansion of farmland. These findings reveal a dynamic feedback loop between land, climate, and hazards, providing critical evidence for targeted land-use zoning, the preservation of natural buffers in climate transition zones, and the integration of climate shift maps into early-warning systems for disaster-prone regions.

Due to the large data volume, the UAV image stitching and matching suffers from high computational cost. The traditional feature extraction algorithms—such as Scale-Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), and Oriented FAST Rotated BRIEF (ORB)—require heavy computation to extract and describe features in high-resolution UAV images. To overcome this issue, You Only Look Once version 3 (YOLOv3) combined with the traditional feature point matching algorithms is utilized to extract descriptive features from the drone dataset of residential areas for roof detection. Unlike the traditional feature extraction algorithms, YOLOv3 performs the feature extraction solely on the proposed candidate regions instead of the entire image, thus the complexity of the image matching is reduced significantly. Then, all the extracted features are fed into Structural Similarity Index Measure (SSIM) to identify the corresponding roof region pair between consecutive image sequences. In addition, the candidate corresponding roof pair by our architecture serves as the coarse matching region pair and limits the search range of features matching to only the detected roof region. This further improves the feature matching consistency and reduces the chances of wrong feature matching. Analytical results show that the proposed method is 13× faster than the traditional image matching methods with comparable performance.

Point cloud classification is an important task in point cloud data analysis. Traditional point cloud classification is conducted primarily on the basis of specific handcrafted features with a specific classifier and is often capable of producing satisfactory results. However, the extraction of crucial handcrafted features hinges on sufficient knowledge of the field and substantial experience. In contrast, while powerful deep learning algorithms possess the ability to learn features automatically, it normally requires complex network architecture and a considerable amount of calculation time to attain better accuracy of classification. In order to combine the advantages of both the methods, in this study, we integrated the handcrafted features, whose benefits were confirmed by previous studies, into a deep learning network, in the hopes of solving the problem of insufficient extraction of specific features and enabling the network to recognise other effective features through automatic learning. This was done to achieve the performance of a complex model by using a simple model and fulfil the application requirements of the remote sensing domain. As indicated by the experimental results, the integration of handcrafted features into the simple and fast-calculating PointNet model could generate a classification result that bore comparison with that generated by a complex network model such as PointNet++ or KPConv.

Many researches have been proved that the most important information of high-dimensional data (e.g. hyperspectral images, HSI) lies in a low-dimensional subspace spanned by some specific bases. Moreover, image pixels belonging to the same class are usually distributed in the same low-dimensional subspace. Therefore, the reduction of dimensionality for classification has become the major issue in hyperspectral image analysis. Sparse representation techniques can be used to process data with sparsity, so it is quite suitable for hyperspectral image analysis. Based on the sparse representation, the paper proposes a method suitable for hyperspectral image classification with learned dictionaries. In the proposed method, the spectral and spatial information are integrated simultaneously into a joint sparse representation in order to increase performances of hyperspectral image classification.

In this study, an accelerating method of oriented FAST and rotated BRIEF combined with principal component analysis (ORB/PCA) is proposed for splicing detection of unmanned aerial vehicle (UAV) images. Compared to traditional scale-invariant feature transform (SIFT) and speeded up robust features (SURF) methods, the proposed ORB/PCA can not only be faster but also produce more accurate. Moreover, in order for the proposed ORB to be effective for image stitching process in near real-time, the Compute Unified Device Architecture (CUDA) application programming interface of graphics processing unit (GPU) is cooperated to speed up the proposed method. Experimental results show that the proposed GPU based ORB/PCA framework is suitable for splicing detection of UAV images in Earth remote sensing. It can improve the image stitching process both in time and accuracy compared to conventional methods.

Pan-sharpening is an image fusion technique of synthesizing a high-resolution multispectral image from a low-resolution multispectral image and a high-resolution panchromatic image. In this paper, a novel pan-sharpening method for remote sensing images has been proposed with sparse representation over learned dictionaries. In the proposed method, the dictionaries are learned only from the high-resolution panchromatic image via the joint learning algorithm, instead of learning from the high-resolution multispectral image which are not available in practice. The sparse coefficients of the panchromatic image and low-resolution panchromatic image are calculated by the orthogonal matching pursuit algorithm. Then, the fused high-resolution multispectral image can be constructed by combining the obtained sparse coefficients and the high-resolution dictionary. The experiment results indicate that the proposed method not only preserve spectral and spatial details of the source images but overcoming the drawbacks of fusion distortion.

The idea of using artificial neural network has been proven useful for hyperspectral image classification. However, the high dimensionality of hyperspectral images usually leads to the failure of constructing an effective neural network classifier. To improve the performance of neural network classifier, wavelet-based feature extraction algorithms can be applied to extract useful features for hyperspectral image classification. However, the extracted features with fixed position and dilation parameters of the wavelets provide insufficient characteristics of spectrum. In this study, wavelet networks which integrates the advantages of wavelet-based feature extraction and neural networks classification is proposed for hyperspectral image classification. Wavelet networks is a kind of feed-forward neural networks using wavelets as activation function. Both the position and the dilation parameters of the wavelets are optimized as well as the weights of the network during the training phase. The value of wavelet networks lies in their capabilities of optimizing network weights and extracting essential features simultaneously for hyperspectral images classification. In this study, the influence of the learning rate and momentum term during the network training phase is presented, and several initialization modes of wavelet networks were used to test the performance of wavelet networks.