Automatic monitoring of vehicles and vessels in satellite imagery
Since 2013, The EU Satellite Centre (SatCen) delivers services and information products for border surveillance to the European Border and Coast Guard Agency, Frontex. This requires continuous improvement of services and capturing new developments in the market in order to eventually introduce them in operations. A daily concern in the Operations Division is dealing with large datasets of very high resolution images, looking for small objects that may be related to illegal migration or cross-border crime. SatCen thus wanted to investigate the potential of AI to help analysts monitor vehicle and vessels in these high resolution satellite images.
Within this project, I developed a pipeline to be able to run deep learning models on this data, automatically dividing these orthomosaic images in smaller chunks with overlap between, in order to deal with objects that lie on the border between different chunks. Using this pipeline, I then evaluated a variety of different single shot detection models for the specific case of vehicle and vessel detection.
References
Vehicle and Vessel Detection on Satellite Imagery: A Comparative Study on Single-Shot Detectors
Tanguy Ophoff, Steven Puttemans, Vasileios Kalogirou, Jean-Philippe Robin, and Toon Goedemé
In this paper, we investigate the feasibility of automatic small object detection, such as vehicles and vessels, in satellite imagery with a spatial resolution between 0.3 and 0.5 m. The main challenges of this task are the small objects, as well as the spread in object sizes, with objects ranging from 5 to a few hundred pixels in length. We first annotated 1500 km2, making sure to have equal amounts of land and water data. On top of this dataset we trained and evaluated four different single-shot object detection networks: YOLOV2, YOLOV3, D-YOLO and YOLT, adjusting the many hyperparameters to achieve maximal accuracy. We performed various experiments to better understand the performance and differences between the models. The best performing model, D-YOLO, reached an average precision of 60% for vehicles and 66% for vessels and can process an image of around 1 Gpx in 14 s. We conclude that these models, if properly tuned, can thus indeed be used to help speed up the workflows of satellite data analysts and to create even bigger datasets, making it possible to train even better models in the future.
@article{ophoff_satellite,author={Ophoff, Tanguy and Puttemans, Steven and Kalogirou, Vasileios and Robin, Jean-Philippe and Goedem{\'e}, Toon},title={Vehicle and Vessel Detection on Satellite Imagery: A Comparative Study on Single-Shot Detectors},journal={Remote Sensing},volume={12},year={2020},number={7},article-number={1217},issn={2072-4292},doi={10.3390/rs12071217},publisher={MDPI},url={https://www.mdpi.com/2072-4292/12/7/1217},}