Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images
| Autor: | Mehmet Soydas, Elif Sertel, Ugur Alganci |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
Computer science
Science end-to-end detection 0211 other engineering and technologies 02 engineering and technology transfer learning single shot multi-box detector (ssd) Convolutional neural network convolutional neural networks (CNNs) remote sensing single shot multi-box detector (SSD) You Look Only Once-v3 (YOLO-v3) Faster RCNN you look only once-v3 (yolo-v3) 0202 electrical engineering electronic engineering information engineering Computer vision 021101 geological & geomatics engineering convolutional neural networks (cnns) business.industry Deep learning Detector Object (computer science) Object detection Visual inspection General Earth and Planetary Sciences 020201 artificial intelligence & image processing Satellite faster rcnn Artificial intelligence business Rotation (mathematics) |
| Zdroj: | Remote Sensing; Volume 12; Issue 3; Pages: 458 Remote Sensing, Vol 12, Iss 3, p 458 (2020) |
| ISSN: | 2072-4292 |
| DOI: | 10.3390/rs12030458 |
| Popis: | Object detection from satellite images has been a challenging problem for many years. With the development of effective deep learning algorithms and advancement in hardware systems, higher accuracies have been achieved in the detection of various objects from very high-resolution (VHR) satellite images. This article provides a comparative evaluation of the state-of-the-art convolutional neural network (CNN)-based object detection models, which are Faster R-CNN, Single Shot Multi-box Detector (SSD), and You Look Only Once-v3 (YOLO-v3), to cope with the limited number of labeled data and to automatically detect airplanes in VHR satellite images. Data augmentation with rotation, rescaling, and cropping was applied on the test images to artificially increase the number of training data from satellite images. Moreover, a non-maximum suppression algorithm (NMS) was introduced at the end of the SSD and YOLO-v3 flows to get rid of the multiple detection occurrences near each detected object in the overlapping areas. The trained networks were applied to five independent VHR test images that cover airports and their surroundings to evaluate their performance objectively. Accuracy assessment results of the test regions proved that Faster R-CNN architecture provided the highest accuracy according to the F1 scores, average precision (AP) metrics, and visual inspection of the results. The YOLO-v3 ranked as second, with a slightly lower performance but providing a balanced trade-off between accuracy and speed. The SSD provided the lowest detection performance, but it was better in object localization. The results were also evaluated in terms of the object size and detection accuracy manner, which proved that large- and medium-sized airplanes were detected with higher accuracy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|