A High Resolution Optical Satellite Image Dataset for Ship Recognition and Some New Baselines

Liu, Zikun; Liu, Yuan; Weng, Lubin; Yang, Yiping

doi:10.5220/0006120603240331

Cited by 484 publications

(278 citation statements)

References 26 publications

Supporting

Mentioning

277

Contrasting

Order By: Relevance

“…With these structures, high performance and high speed could be achieved in real-time applications, such as object detection from a video stream. Additionally, object proposal approaches have become more widely used in remote sensing applications, with improvements in speed and performance [29][30][31][32][33]. Detection by producing an object proposal achieved successful results, but there is a trade-off between the detection performance and processing speed according to the number of proposals produced.…”

Section: Introductionmentioning

confidence: 99%

Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images

2020

View full text Add to dashboard Cite

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.

show abstract

Section: Introductionmentioning

confidence: 99%

Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images

2020

View full text Add to dashboard Cite

show abstract

“…Google Earth/Microsoft Virtual Earth images are freely available for anyone to view from the application imagery and for personal use. They offer spectral, texture, geometry and other characteristics for analysis, and, as stated by Liu et al (2017), experiments on images derived from Google Earth are important references to practical applications with optical data directly collected from spaceborne devices. But since these images are pre-processed for visualisation purposes only, they have relatively poor spectral information (with red, green and blue bands), and also large inconsistent radiometric distortions (Yu and Gong, 2012).…”

Section: Discussionmentioning

confidence: 99%

“…These methods have high performance and provide good length estimation, but at the same time, they require high knowledge of the operator. For the needs of training data, Liu et al (2017) have compiled HRSC2016 (High Resolution Ship Collection 2016), a set of reference data for vessel recognition from publicly available high-resolution imagery for both at-sea and inshore vessels. The vessel models are organised into a tree structure with three levels (vessel class, category and type).…”

Section: Discussionmentioning

confidence: 99%

“…The most calculated geometrical feature for discrimination is the size, i.e. the length and the width of a vessel candidate (Buck et al, 2008, Heiselberg, 2016, Lin et al, 2012, Liu et al, 2014, Liu et al, 2017, Pegler et al, 2003, Willhauck et al, 2005). Also area (Deng et al, 2013, Dong et al, 2013, Partsinevelos and Miliaresis, 2014, Tao et al, 2009) and length-to-width ratio are quite common for discrimination (Chen et al, 2009, Guang et al, 2011, Hong et al, 2007, Jubelin and Khenchaf, 2014, Wang et al, 2011, Xiaoyang et al, 2016).…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Vessel detection and classification from spaceborne optical images: A literature survey

Kanjir

Greidanus

Oštir

2018

Remote Sensing of Environment

249

147

View full text Add to dashboard Cite

show abstract

“…We choose two datasets to experiment with our framework in high-resolution aerial images. One was the public multiclass ship dataset HRSC2016 [51], and the other was our constructed dataset Multiclass Warship Collection (MWC).…”

Section: Datasetsmentioning

confidence: 99%

Towards Automated Ship Detection and Category Recognition from High-Resolution Aerial Images

et al. 2019

View full text Add to dashboard Cite

Ship category classification in high-resolution aerial images has attracted great interest in applications such as maritime security, naval construction, and port management. However, the applications of previous methods were mainly limited by the following issues: (i) The existing ship category classification methods were mainly to classify on accurately-cropped image patches. This is unsatisfactory for the results of the existing methods in practical applications, because the location of the ship in the patch obtained by the object detection varies greatly. (ii) The factors such as target scale variations and class imbalance have a great influence on the performance of ship category classification. Aiming at the issues above, we propose a novel ship detection and category classification framework. The category classification is based on accurate location. The detection network can generate more precise rotated bounding boxes in large-scale aerial images by introducing a novel Sequence Local Context (SLC) module. Besides, three different ship category classification networks are proposed to eliminate the effect of scale variations, and the Spatial Transform Crop (STC) operation is used to get aligned image patches. Whatever the problem of insufficient samples or class imbalance have, the Proposals Simulation Generator (PSG) is considered to handle this properly. Most remarkably, the state-of-the-art performance of our framework is demonstrated by experiments based on the 19-class ship dataset HRSC2016 and our multiclass warship dataset.

show abstract

A High Resolution Optical Satellite Image Dataset for Ship Recognition and Some New Baselines

Cited by 484 publications

References 26 publications

Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images

Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images

Vessel detection and classification from spaceborne optical images: A literature survey

Towards Automated Ship Detection and Category Recognition from High-Resolution Aerial Images

Contact Info

Product

Resources

About