The RADARSAT-2 satellite enables new capabilities that complement optical satellite data such as broad area change alerting and all season/weather target monitoring. The RADARSAT-2 mission provides a unique imaging mode called ExtraFine which has 5m resolution and a scene size of 125 km x 135 km so it can provide complete overlapping coverage, even at the equator. When radar imagery is collected in exact repeat geometry, we can automate the extraction of change information over vast areas and create routine change detection solutions to monitor changes such as construction or port activity.

ExtraFine Archive of Over 3 Billion Square Km

The ExtraFine archive available to GBDX spans from May 2014 to present, and consists primarily of exact repeat geometry collects on a 24 day cycle in ascending and descending collections. As of January 2017, there are approximately 3 billion square kilometers of ExtraFine imagery in the archive (~200,000 scenes), most of which are over land and ideal for broad area change monitoring. The figure to the right shows a density map of the distribution of the ExtraFine archive which primarily consists of routine collections over cloud covered parts of the Earth, as well as, defense and intelligence priority areas.

Exact Repeat Geometry Enables Automated Big Data Analysis

When Synthetic Aperture Radar (SAR) imagery is collected in the exact repeat geometry, noise and distortions can be minimized. This simplifies the processing and enables easier automation of change detection processing. Further, this enables advanced image processes such as Multi-Temporal Filtering (MTF) to remove speckle and enhance the quality of the imagery, making change detection more effective. The figure below provides an example of the benefit of working with exact repeat geometry stacks where the left hand image is un-filtered and the right hand image has MTF applied based on a stack of 5 images. In the example below you can see the clear difference in the image quality.

Change Products and How to Interpret Them

The MDA change detection tool employs a stack-based Amplitude Change Detection (ACD) methodology. The algorithm pre-processes the imagery stack to remove speckle, atmospheric effects, and fine registers the imagery to a sub-pixel level, enabling the image to image for comparison. The intensity of the aligned pixels varies as the composition of the scene varies. For example, if a truck is parked in the scene or a new building is built, this would have a higher radar reflectance and present as a brighter pixel. The variation in pixel brightness can be filtered to detect persistent changes in a scene. The figure below shows how this can be examined on a pixel scale. Further analysis on grouping of changes can enhance the change result to filter only relevant changes to the target application. The ACD algorithm takes threshold and morphological inputs to help filter out irrelevant changes for the application.

Using this methodology we are able to create a time series of changes as shown in the figure to the right. This example shows a series of RADARSAT-2 images and comparable optical images where available. The blue pixels represent new brighter targets and the red pixels represent new darker areas (“blue is new, red has fled”). From this time series, it is evident that construction is ongoing with blue activity representing machinery, foundation work, new building activity, and red indicating creation of low reflectivity areas such as parking lots and roads.

Examples: Monitoring The Port of Murmansk

The following ACD example over Murmansk demonstrates the utility of RADARSAT-2 images for monitoring harbor activity. The same images used in this case for harbor activity monitoring can be employed to detect new construction, roads, and other infrastructure changes.

Murmansk is a town located in the extreme northwest part of Russia at 68.9 degrees North. This region contains several naval facilities such as a nuclear submarines base at Gadzhiyevo and a naval missile depot in Severomorsk. This region is characterized by high cloud cover and prolonged periods of complete darkness in the winter. The figure (right) shows the area of interest over Murmansk and the Russian fleet ports of Olenya and Severomosk. The Cyan box is the ExtraFine scene foot print (125x125km). This AOI represents a stack of imagery collected every 24 days. More frequent collects are possible; imagery can be acquired up to every 2-3 days depending on latitude.

The figure below shows an example output from the end-to-end ACD task. The top two images are the before/after images from 2016-06-06 and 2016-06-30, and the bottom image is the output of ACD overlaid on a RADARSAT-2 image. In this example the ACD task detected changes in vessel positions and marked them as blue or red (blue is new or arrived, red has fled or departed). Note that a large ship (Ship 1) has departed from Olenya Bay. One vessel (ship 2) appears as new, but is likely a repositioning of another vessel from the east side of the harbor (also circled in yellow). This example demonstrates how RADARSAT-2 can be used for initial target detection and site monitoring for cueing of a higher-resolution optical and/or radar acquisitions to identify the targets and provide detailed site context.

Olenya Bay Port Example

ACD XF results between June 6, 2016 and June 30, 2016 over Murmansk, AOI 1: Olenya Bay. The top images are the two images from the Ascending XFW03 stack used for ACD. The bottom image is the RB CD map. The area extent of the image chip is 2.3 km x 1.3 km (range x azimuth).  Ship 1 departure is probably the same ship identified below in a WorldView image dated 2016-05-21 based on the ship superstructure location and ship length.

 

Severomosk Bay Port Example

ACD XF results between April 19, 2016 and May 13, 2016 over Murmansk, AOI 2: Severomorsk. The top images are the two images from the Ascending XFW03 stack used for ACD. The bottom image is the RB CD map. The area extent of the image chip is 4.4 km x 2.3 km (range x azimuth).

Interested in how you can use RADARSAT-2 on GBDX to address your location intelligence challenges? Contact Us.

RADARSAT-2 Data and Products © MacDonald, Dettwiler and Associates Ltd. – All Rights Reserved. RADARSAT is an official mark of the Canadian Space Agency.

Value Added Products (VAPs) created by MDA GSI
© MDA Geospatial Services Inc.– All Rights Reserved.