Airports have a profound impact on our lives, and uncovering their distribution around the world has great significance for research and development. However, existing airport databases are incomplete and have a high cost of updating. Thus, a fast and automatic worldwide airport detection method can be of significance for global airport detection at regular intervals. However, previous airport detection studies are usually based on single remote sensing (RS) imagery, which seems an overwhelming burden for worldwide airport detection with traversal searching. Thus, we propose a hierarchical airport detection method consisting of broad-scale extraction of worldwide candidate airport regions based on spatial analysis of released RS products, including impervious surfaces from FROM-GLC10 (fine resolution observation and monitoring of global land cover 10) product, building distribution from OSMs (open street maps) and digital surface model from AW3D30 (ALOS World 3D—30 m). Moreover, narrow-scale aircraft detection was initially conducted by the Faster R-CNN (regional-convolutional neural networks) deep learning method. To avoid overestimation of background regions by Faster R-CNN, a second CNN classifier is used to refine the class labeling with negative samples. Specifically, our research focuses on target airports with at least 2 km length in three experimental regions. Results show that spatial analysis reduced the possible regions to 0.56% of the total area of 75,691 km². The initial aircraft detection by Faster R-CNN had a mean user’s accuracy of 88.90% and ensured that all the aircrafts could be detected. Then, by introducing the CNN reclassifier, the user’s accuracy of aircraft detection was significantly increased to 94.21%. Finally, through an experienced threshold of aircraft number, 19 of the total 20 airports were detected correctly. Our results reveal the overall workflow is reliable for automatic and rapid airport detection around the world with the help of released RS products. This research promotes the application and progression of deep learning.

Airports are important transportation hubs, but their locations searched on open databases are not reliable. Manual verification of these locations is time-consuming and labor-intensive, so that a rapid and automated validation of their spatial location is necessary. In this study, three datasets of global airports were collected and fused into one dataset through coordinate and name matching. The fused dataset contained 46,290 airport (with runway) records. Then, we downloaded the remote sensing images of these airports from Google Earth. To determine whether there were airports in these images, we proposed a process framework. In this framework, we used a multi-scale runway detector based on YOLO-v3 to initially detect the airport runway, then used a re-classifier based on ResNet-101 to improve the accuracy of the initial detection results, and gave a comprehensive result score. The accuracy of this process framework and the airport recall rate on the test dataset reached 95.4% and 95.6%, respectively. The framework was applied to airport locations around the world. When the threshold of the result score was set to 0.65, 29259 airport records passed the verification. In addition, we manually verified the application results. The accuracy of the process framework reached 91%, while its speed was 15 times faster than that of the manual verification. The results showed that the entire process framework can quickly and reliably help verify the spatial locations of airports worldwide and provide processing ideas for the validation of the spatial locations of other remote sensing objects.

Terrestrial laser scanning (TLS) has been used widely in topographic deformation monitoring, and the issue of deformation information extraction from multi-temporal point cloud data has high importance. Here, a method is developed for the extraction of topographic deformation in a forested area based on the 3D spatial information of individual trees. Initially, ground points are removed by using a local topographic surface fitting method, then individual trees are segmented using the PTrees algorithm; finally, the motion of individual trees is computed with a combination of global and local rigid-body transformations. Experimental validation results show that the proposed method is reliable to within a 0.12° maximum tilt error and 8.7 mm maximum displacement error. In addition, some inferences were drawn based on the original topography and the topographic deformation monitoring results. For example, a mining work-face was located beyond the eastern part of the study area, and trees growing in the middle part of this region of interest may face a survival challenge.