The timely and proper rehabilitation of damaged roads is essential for road maintenance, and an effective method to detect road surface distress with high efficiency and low cost is urgently needed. Meanwhile, unmanned aerial vehicles (UAVs), with the advantages of high flexibility, low cost, and easy maneuverability, are a new fascinating choice for road condition monitoring. In this paper, road images from UAV oblique photogrammetry are used to reconstruct road three-dimensional (3D) models, from which road pavement distress is automatically detected and the corresponding dimensions are extracted using the developed algorithm. Compared with a field survey, the detection result presents a high precision with an error of around 1 cm in the height dimension for most cases, demonstrating the potential of the proposed method for future engineering practice.

Floods, storms and hurricanes are devastating for human life and agricultural cropland. Near-real-time (NRT) discharge estimation is crucial to avoid the damages from flood disasters. The key input for the discharge estimation is precipitation. Directly using the ground stations to measure precipitation is not efficient, especially during a severe rainstorm, because precipitation varies even in the same region. This uncertainty might result in much less robust flood discharge estimation and forecasting models. The use of satellite precipitation products (SPPs) provides a larger area of coverage of rainstorms and a higher frequency of precipitation data compared to using the ground stations. In this paper, based on SPPs, a new NRT flood forecasting approach is proposed to reduce the time of the emergency response to flood disasters to minimize disaster damage. The proposed method allows us to forecast floods using a discharge hydrograph and to use the results to map flood extent by introducing SPPs into the rainfall–runoff model. In this study, we first evaluated the capacity of SPPs to estimate flood discharge and their accuracy in flood extent mapping. Two high temporal resolution SPPs were compared, integrated multi-satellite retrievals for global precipitation measurement (IMERG) and tropical rainfall measurement mission multi-satellite precipitation analysis (TMPA). The two products are evaluated over the Ottawa watershed in Canada during the period from 10 April 2017 to 10 May 2017. With TMPA, the results showed that the difference between the observed and modeled discharges was significant with a Nash–Sutcliffe efficiency (NSE) of −0.9241 and an adapted NSE (ANSE) of −1.0048 under high flow conditions. The TMPA-based model did not reproduce the shape of the observed hydrographs. However, with IMERG, the difference between the observed and modeled discharges was improved with an NSE equal to 0.80387 and an ANSE of 0.82874. Also, the IMERG-based model could reproduce the shape of the observed hydrographs, mainly under high flow conditions. Since IMERG products provide better accuracy, they were used for flood extent mapping in this study. Flood mapping results showed that the error was mostly within one pixel compared with the observed flood benchmark data of the Ottawa River acquired by RadarSat-2 during the flood event. The newly developed flood forecasting approach based on SPPs offers a solution for flood disaster management for poorly or totally ungauged watersheds regarding precipitation measurement. These findings could be referred to by others for NRT flood forecasting research and applications.

Building information is an essential part of geographic information system (GIS) applications in urban planning and management. However, it changes rapidly with economic growth. Unmanned aerial vehicles (UAV)-based photogrammetry works well in this situation with its advantages of quick and high-resolution data updating. In this paper, in order to improve building extraction accuracy in complex areas where buildings are characterized by various patterns, complex structures, and unique styles, we present a framework which applies deep learning (DL) semantic segmentation to UAV images with digital surface model (DSM) and visible-band difference vegetation index (VDVI). The results show that extraction accuracy improves. The combination of red, green, blue (RGB) and VDVI bands (RGBVI) can effectively distinguish the building area and vegetation. The application of RGB with DSM bands (RGBD) helps separate buildings from ground objects. The combination of RGB, DSM, and VDVI bands (RGBDVI) can identify small buildings which are usually not high and covered partly by tree branches. The proposed method is further applied to an open standard dataset to evaluate its robustness and results indicate an increased overall accuracy from RGB only (93%) to RGBD (97%).

Using deep learning semantic segmentation for land use extraction is the most challenging problem in medium spatial resolution imagery. This is because of the deep convolution layer and multiple levels of deep steps of the baseline network, which can cause a degradation problem in small land use features. In this paper, a deep learning semantic segmentation algorithm which comprises an adjustment network architecture (LoopNet) and land use dataset is proposed for automatic land use classification using Landsat 8 imagery. The experimental results illustrate that deep learning semantic segmentation using the baseline network (SegNet, U-Net) outperforms pixel-based machine learning algorithms (MLE, SVM, RF) for land use classification. Furthermore, the LoopNet network, which comprises a convolutional loop and convolutional block, is superior to other baseline networks (SegNet, U-Net, PSPnet) and improvement networks (ResU-Net, DeeplabV3+, U-Net++), with 89.84% overall accuracy and good segmentation results. The evaluation of multispectral bands in the land use dataset demonstrates that Band 5 has good performance in terms of extraction accuracy, with 83.91% overall accuracy. Furthermore, the combination of different spectral bands (Band 1–Band 7) achieved the highest accuracy result (89.84%) compared to individual bands. These results indicate the effectiveness of LoopNet and multispectral bands for land use classification using Landsat 8 imagery.

Buildings along riverbanks are likely to be affected by rising water levels, therefore the acquisition of accurate building information has great importance not only for riverbank environmental protection but also for dealing with emergency cases like flooding. UAV-based photographs are flexible and cloud-free compared to satellite images and can provide very high-resolution images up to centimeter level, while there exist great challenges in quickly and accurately detecting and extracting building from UAV images because there are usually too many details and distortions on UAV images. In this paper, a deep learning (DL)-based approach is proposed for more accurately extracting building information, in which the network architecture, SegNet, is used in the semantic segmentation after the network training on a completely labeled UAV image dataset covering multi-dimension urban settlement appearances along a riverbank area in Chongqing. The experiment results show that an excellent performance has been obtained in the detection of buildings from untrained locations with an average overall accuracy more than 90%. To verify the generality and advantage of the proposed method, the procedure is further evaluated by training and testing with another two open standard datasets which have a variety of building patterns and styles, and the final overall accuracies of building extraction are more than 93% and 95%, respectively.

Multilabel remote sensing (RS) image annotation is a challenging and time-consuming task that requires a considerable amount of expert knowledge. Most existing RS image annotation methods are based on handcrafted features and require multistage processes that are not sufficiently efficient and effective. An RS image can be assigned with a single label at the scene level to depict the overall understanding of the scene and with multiple labels at the object level to represent the major components. The multiple labels can be used as supervised information for annotation, whereas the single label can be used as additional information to exploit the scene-level similarity relationships. By exploiting the dual-level semantic concepts, we propose an end-to-end deep learning framework for object-level multilabel annotation of RS images. The proposed framework consists of a shared convolutional neural network for discriminative feature learning, a classification branch for multilabel annotation and an embedding branch for preserving the scene-level similarity relationships. In the classification branch, an attention mechanism is introduced to generate attention-aware features, and skip-layer connections are incorporated to combine information from multiple layers. The philosophy of the embedding branch is that images with the same scene-level semantic concepts should have similar visual representations. The proposed method adopts the binary cross-entropy loss for classification and the triplet loss for image embedding learning. The evaluations on three multilabel RS image data sets demonstrate the effectiveness and superiority of the proposed method in comparison with the state-of-the-art methods.