L. W. Mdakane

Very deep learning for ship discrimination in Synthetic Aperture Radar imagery

Efficient and effective ship discrimination across multiple Synthetic Aperture Radar sensors is becoming more important as access to SAR data becomes more widespread. A flexible means of separating ships from sea is ideal and can be accomplished using machine learning. Newer, advanced deep learning techniques offer a unique solution but traditionally require a large dataset to train effectively. Highway Networks allow for very deep networks that can be trained using the smaller datasets typical in SAR-based ship detection. A flexible network configuration is possible within Highway Networks due to an adaptive gating mechanism which prevents gradient decay across many layers. This paper presents a very deep High Network configuration as a ship discrimination stage for SAR ship detection. It also presents a three-class SAR dataset that allows for more meaningful analysis of ship discrimination performances. The proposed method was tested on a this SAR dataset and had the highest mean accuracy of all methods tested at 96.67%. The proposed ship discrimination method also provides improved false positive classification compared to the other methods tested.

Using an active contour method to detect bilge dumps from SAR imagery

An automatic approach to detect bilge dumping in synthetic aperture radar (SAR) images over Southern African oceans is proposed. The approach uses a threshold-based algorithm and a region-based active contour model (ACM) algorithm to achieve an efficient bilge dump detection tool. A threshold method was used to detect areas with a high bilge dump probability while the ACM method is used to get closed contours of potential bilge dumps. A discrimination step was implemented to detect linear features as potential bilge dump events. The automated approach was investigated using SENTINEL-1A and ENVISAT Advanced Synthetic Aperture Radar (ASAR) where the proposed approach showed promising results.

Ship detection in Sentinel-1 imagery using the H-dome transformation

The ability to reliably detect ships within the ocean is one of the core capabilities of Maritime Domain Awareness. Constant improvements are pursued in both computational speed and reliable responses to events by exploring new ship detection methods. One such improvement is the availability of Sentinel-1 imagery. Owing to the fact that ships are considered locally bright objects, we propose using the H-dome transform to process the SAR image and improve ship detectability. The method was tested against two Sentinel-1 images in both HH and HV polarizations containing 82 ships. The method improved the false alarm rates when compared to the conventional cell-averaging Constant False Alarm Rate method at a minor reduction in detection accuracy.

Bilge dump detection from SAR imagery using local binary patterns

Accidental or deliberate bilge dumping presents a major threat to the sea ecosystem. We present a semi automatic approach to detect bilge dumping in synthetic aperture radar (SAR) images. The approach consist of three main parts. Firstly, areas with high probability of being bilge dumps are detected using Local Binary Patterns (LBP) with an adaptive threshold. Secondly, features are extracted from the detected dark spots and lastly, the features are analysed using bilge dump database to discriminate dark spot as bilge or not bilge. The automated approach was investigated on nine visually inspected images of SENTINEL 1A and ENVISAT Advanced Synthetic Aperture Radar (ASAR) images. The performance was measured by comparing the number of detected bilge dumps using the automated approach with the visually detected database. The automated detection approach showed to be a good alternative of the labour intensive manual inspection of bilge dumps, particularly for large ocean area monitoring.

Image segmentation-based oil slick detection using sar radarsat-2 OSVN maritime data

Oil spills present a major threat to the sea ecosystem and thus need to be monitored on a regular basis. Synthetic Aperture Radar (SAR) data is well known for ocean monitoring capabilities. SENTINEL 1 (SEN1) extra wide (EW) mode data and RADARSAT-2 (RS2) Maritime Satellite Surveillance Radar (MSSR) modes have been developed to further improve ocean surveillance. This data can monitor large areas (400 km for SEN1 EW and over 500 km for RS2 OSVN), with a finer resolution. These modes enable improved oil slick detection (including ship detection to identify the source) performance while reducing the number of needed scenes. Numerous studies have been proposed for SEN1 data due to its free access while less work has been done on oil spill detection methods using the RS2 OSVN data. In this paper, we evaluated a segmentation-based method on RS2 OSVN data.

An Image-Segmentation-Based Framework to Detect Oil Slicks From Moving Vessels in the Southern African Oceans Using SAR Imagery

Oil slick events caused due to bilge leakage/dumps from ships and from other anthropogenic sources pose a threat to the aquatic ecosystem and need to be monitored on a regular basis. An automatic image-segmentation-based framework to detect oil slick from moving vessels using spaceborne synthetic aperture radar (SAR) images over Southern African oceans was proposed. The study uses an automated threshold-based algorithm and a region-based algorithm to achieve a more efficient oil slick detection. The proposed framework consisted of two parts: First, a threshold-based method was used to detect areas with a high oil slick probability; second, a region-based method was used to extract the full extent of the detected oil slick. The proposed framework was tested on both real SAR and synthetic SAR images and was robust to intensity variations, weak boundaries, and was also more computationally efficient when compared to the region-based method without the threshold-based input.

Ships as salient objects in Synthetic Aperture Radar imagery

The widespread access to Synthetic Aperture Radar data has created a need for more precise ship extraction, specifically in low-to-medium resolution imagery. While Synthetic Aperture Radar pixel resolution is improving for a large swaths, information about ships from within the Synthetic Aperture Radar intensity imagery is still sparse. Ships that are a few pixels across provide little information for classification and even less when improperly extracted. This paper presents a novel perspective on ships in Synthetic Aperture Radar imagery by viewing them as visually salient objects. The paper introduces common methods of ship object extraction and demonstrates how salient object mapping can improve the accuracy of extracted ships in Synthetic Aperture Radar imagery, providing better representation of ship objects. The Frequency-tuned and Spectral Residual Saliency Maps methods were tested against a unique dataset with ground truth information and were shown to have the best performance amongst all the conventional methods tested using six performance metrics.

A CA-CFAR and localized wavelet ship detector for Sentinel-1 imagery

The Maritime Domain Awareness initiative seeks to constantly improve the ways in which maritime information is collected. With the recent release of free Sentinel-1 imagery to the public, monitoring the maritime environment has become a more affordable. Using the basis of a cell-averaging constant false alarm rate prescreening method as input, this paper presents a novel method for detecting ships within Synthetic Aperture Radar imagery using a Gabor wavelet correlator. The method proposed allows for any configuration of the filter bank and prevents false detections by processing possible targets at a local scale. The method was tested against two Sentinel-1 images in both HH and HV polarization with a total of 82 ships. The method provided significantly improved FAR over the conventional CA-CFAR method at the cost of slightly worse detection accuracies in some cases.