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.

Evaluating land and runoff variations caused by urbanization is crucial to ensure the safety of people living in highly developed areas. Based on spatial scales, runoff analysis involves different methods associated with the interpretation of land cover and land use, the application of hydrological models, and the consideration of flood mitigation measures. Most studies have focused on analyzing the phenomenon on a certain scale by using a single data source and a specific model without discussing mutual influences. In this study, the runoff changes caused by urbanization are assessed and cross-analyzed on three sizes of study areas in the Zhuoshui River Basin in Taiwan, including basin (large), watershed (medium), and city (small) scales. The results demonstrate that, on the basin scale, land-cover changes interpreted from satellite images are very helpful for identifying the watersheds with urbanization hotspots that might have larger runoff outputs. However, on the watershed scale, the resolution of the land-cover data is too low, and land-cover data should be replaced by investigated land-use data for sophisticated hydrological modeling. The mixed usage of land-cover and land-use data is not recommended because large discrepancies occur when determining hydrological parameters for runoff simulation. According to present and future land-use scenarios, the influence of urbanization on runoff is simulated by HEC-1 and SWMM on watershed and city scales, respectively. The results of both models are in agreement and show that runoff peaks will obviously increase as a result of urbanization from 2008 to 2030. For low return periods, the increase in runoff as a result of urbanization is more significant and the city’s contribution to runoff is much larger than its area. Through statistical regression, the watershed runoff simulated by HEC-1 can be perfectly predicted by the city runoff simulated by SWMM in combination with other land/rainfall parameters. On the city scale, the installation of LID satisfactorily reduces the runoff peaks to pre-urbanization levels for low return periods, but the effects of LID are not as positive and are debatable for higher return periods. These findings can be used to realize the applicability and limitations of different approaches for analyzing and mitigating urbanization-induced runoff in the process of constructing a sponge city.