Ship Classification Based on MSHOG Feature and Task-Driven Dictionary Learning with Structured Incoherent Constraints in SAR Images

Lin, Hung‐Mo; Song, Shaoxu; Yang, Jian

doi:10.3390/rs10020190

Cited by 42 publications

(19 citation statements)

References 40 publications

(66 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another limitation of HSD is that it could not help identify changes if they were missed by both methods no matter how to calculate the weights. In this situation, spectral characteristics were not enough to find changes, and more features like texture, e.g., grey-level co-occurrence matrix (GLCM) [46] and histogram of oriented gradients (HOG) [47], should be used.…”

Section: (B) Weights For Cdsv (A) Weights For Cdssmentioning

confidence: 99%

A Novel Approach to Unsupervised Change Detection Based on Hybrid Spectral Difference

Wang

Zhan

et al. 2018

Remote Sensing

View full text Add to dashboard Cite

The most commonly used features in unsupervised change detection are spectral characteristics. Traditional methods describe the degree of the change between two pixels by quantifying the difference in spectral values or spectral shapes (spectral curve shapes). However, traditional methods based on variation in spectral shapes tend to miss the change between two pixels if their spectral curves are close to flat; and traditional methods based on variation in spectral values tend to miss the change between two pixels if their values are low (dark objects). To inhibit the weaknesses of traditional methods, a novel approach to unsupervised change detection based on hybrid spectral difference (HSD) is proposed which combines the difference between spectral values and spectral shapes. First, a new method referred to as change detection based on spectral shapes (CDSS) is proposed that fuses the difference images produced by the spectral correlation mapper (SCM) and spectral gradient difference (SGD) in order to describe the variation in spectral shapes. Second, a method called change detection based on spectral values (CDSV), computing the Euclidean distance between two spectral vectors, is used to obtain a difference image based on the variation in spectral values. Then, the credibility of CDSS and CDSV for every pixel is calculated to describe how appropriate these two methods are for detecting the change. Finally, the difference images produced by CDSS and CDSV are fused with the corresponding credibility to generate the hybrid spectral difference image. Two experiments were carried out on worldview-2/3 and Landsat-7 Enhanced Thematic Mapper Plus (ETM+) datasets, and both qualitative and quantitative results indicated that HSD had superior capabilities of change detection compared with standard change vector analysis (CVA), SCM, SGD and multivariate alteration detection (MAD). The accuracy of CDSS is higher than CDSV in case-1 but lower in case-2 and, compared to the higher one, the overall accuracy and the kappa coefficient of HSD improved by 3.45% and 6.92%, respectively, in the first experiment, and by 1.66% and 3.31%, respectively, in the second experiment. The omission rate dropped by approx. 4.4% in both tests.

show abstract

Section: (B) Weights For Cdsv (A) Weights For Cdssmentioning

confidence: 99%

A Novel Approach to Unsupervised Change Detection Based on Hybrid Spectral Difference

Wang

Zhan

et al. 2018

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…More recently, [16] classified cargo, tanker, windmill, platform, and harbor structure, with an overall accuracy of 94% using a Convolutional Neural Network (CNN). [17] also reported high classification accuracy (98%) with TerraSAR-X images (1 m resolution), using features obtained from Histogram of Oriented Gradients (HOG) combined with Manifold learning for dimensionality reduction, and a Task-Driven Dictionary Learning framework.…”

Section: Introductionmentioning

confidence: 99%

Maritime Vessel Classification to Monitor Fisheries with SAR: Demonstration in the North Sea

2019

View full text Add to dashboard Cite

Integration of methods based on satellite remote sensing into current maritime monitoring strategies could help tackle the problem of global overfishing. Operational software is now available to perform vessel detection on satellite imagery, but research on vessel classification has mainly focused on bulk carriers, container ships, and oil tankers, using high-resolution commercial Synthetic Aperture Radar (SAR) imagery. Here, we present a method based on Random Forest (RF) to distinguish fishing and non-fishing vessels, and apply it to an area in the North Sea. The RF classifier takes as input the vessel’s length, longitude, and latitude, its distance to the nearest shore, and the time of the measurement (am or pm). The classifier is trained and tested on data from the Automatic Identification System (AIS). The overall classification accuracy is 91%, but the precision for the fishing class is only 58% because of specific regions in the study area where activities of fishing and non-fishing vessels overlap. We then apply the classifier to a collection of vessel detections obtained by applying the Search for Unidentified Maritime Objects (SUMO) vessel detector to the 2017 Sentinel-1 SAR images of the North Sea. The trend in our monthly fishing-vessel count agrees with data from Global Fishing Watch on fishing-vessel presence. These initial results suggest that our approach could help monitor intensification or reduction of fishing activity, which is critical in the context of the global overfishing problem.

show abstract

“…In ref. [5] a new method is proposed to address ship classification in SAR imagery that is based on the combination of an improved SAR-HOG method and a manifold learning methodology to reduce the dimensionality of the problem. Experiments undertaken on actual TerraSAR-X scenes confirm the soundness of the proposed rationale.…”

mentioning

confidence: 99%

Editorial for the Special Issue “Remote Sensing of Target Detection in Marine Environment”

2019

View full text Add to dashboard Cite

Remote sensing is a powerful tool used to obtain an unprecedented amount of information about the ocean from a distance, usually from satellites or aircrafts. Measurements collected by active and passive remote sensing instruments can be used for both marine and maritime applications. They allow monitoring of vast areas of the Earth that are difficult to access and sample using traditional methods. Within this context, the observation of targets at sea, e.g.; man-made targets (ships or oil/gas rigs/platforms and wind turbines) and natural targets (icebergs, surfactants, etc.) is nowadays a very hot-topic in the field of global monitoring of environment and security.Among the remote sensing tools, Synthetic Aperture Radar (SAR) is one of the most used since it allows all-day and almost all-weather observations with a moderate-to-fine spatial resolution. SAR imagery gives the possibility to overcome the limits of maritime patrol allowing non-cooperative all-day target surveillance, over wide regions and under almost all-weather conditions. An increasing number of SAR satellites have become available since the early 1990s. This unprecedented development in SAR sensors requires the definition of new techniques and algorithms to detect marine targets, as well as in the assessment of existing methods. Hence, although there is a great deal of literature that concerns SAR methods for detecting targets at sea, there is still room to improve both models and methods.Within this context, methods and models are proposed in this Special Issue (SI) to exploit single-and multi-polarization SAR measurements for observation of targets at sea, with special interest for vessels.In ref.[1] the complex coherence between the two channels of the dual-polarized Sentinel-1 SAR imagery is exploited to detect ships in a timely manner using the entire image. The proposed rationale is verified against real SAR imagery that includes a large number of vessels of different sizes. In addition, Automatic Identification System (AIS) data are used for an extensive and large-scale cross-comparison with the SAR-based detections. Experimental results clearly point out the remarkable detection rate of the SAR-based method and its complementarity with respect to AIS information.In ref.[2] fishing vessels detection is addressed with reference to the problem of global overfishing. A method is here developed to deal with the discrimination between fishing and non-fishing vessels and a showcase is presented. The method, which is based on the Random Forest (RF) classifier, takes as input the vessel's length, longitude, and latitude, its distance to the nearest shore, and the time of the measurement (am or pm). The classifier is trained and tested on data from the AIS and Sentinel-1 SAR imagery that refers to the North Sea.In ref.[3] a system to monitor illegal fishing activities is proposed. The approach, which is developed in the framework of a cooperation between the Plymouth Marine Laboratory (PML) and

show abstract

Ship Classification Based on MSHOG Feature and Task-Driven Dictionary Learning with Structured Incoherent Constraints in SAR Images

Cited by 42 publications

References 40 publications

A Novel Approach to Unsupervised Change Detection Based on Hybrid Spectral Difference

A Novel Approach to Unsupervised Change Detection Based on Hybrid Spectral Difference

Maritime Vessel Classification to Monitor Fisheries with SAR: Demonstration in the North Sea

Editorial for the Special Issue “Remote Sensing of Target Detection in Marine Environment”

Contact Info

Product

Resources

About