InternetSAR.org

[RiverAux]

http://internetsar.org/

Looks like the google earth effort to find Fosset has relaunched and started a second search in Canada. 

I wonder who is paying for the imagery for the 2nd search.  Also, who is evaluating the leads now. 

So, public affairs officers might want to be prepared to answer questions about why CAP isn't using this resource in the future. 

[Eclipse]

There's been a far number of articles that indicated the mob efforts to analyze the photos for Fossett resulted in a lot of background noise and no results.

[sardak]

The person hosting this Internet search site also has a blog.  Here is an entry from him regarding using the Internet to revolutionize SAR.
Fossett Search Blog

And here is discussion from another SAR lister about another technology they used on this Canadian search.
***********
Recently a number of ground SAR teams in British Columbia were asked to assist in the search for a missing Cessna aircraft, that had gone missing flying 300 miles from Revelstoke, BC, in the Selkirk Mountains, to Qualicum Beach on Vancouver Island. The air-search of this corridor had continued for many days without success. Ground SAR was asked to examine an area near Merritt, BC where the aircraft was believed to have 'dropped off the radar'. This was a high plateau area, 4-5,000ft in elevation, cut through with numerous deep valleys. The region below this possible 'PLS' was intially searched by ground SAR teams driving quads along old access roads. When this searching proved unsuccessful we decided to try a different approach.

Using the aircraft performance characteristics, the effects of wind on the aircraft and an analysis of the terrain profiles beneath the aircraft, we created a 360-degree calculated 'Crash Map' of potential crash-landing (impact) locations for the aircraft, based on the assumptions of an expected Point Last Seen and the aircraft losing power for some reason. The calculated crash map for this scenario, approximately 17 miles in radius, very dramatically reduced the size of the search area, which was then reduced even further to only the 'heights of land' that would block the descent of the un-powered aircraft.

With this calculated 'crash map' air resources and motorized ground SAR teams were able to focus on these high probability potential impact sites, within this very much smaller, targeted, search area.

This Cessna air-search, which had been proceeding for many days, was wound down shortly afterwards, but the same approach was applied, informally, when a helicopter medivac flight recently went missing on a 155 mile flight from Cordova to Anchorage, Alaska. A crash map was calculated for this incident, which determined a potential splashdown point in Prince William Sound. A body and a helicopter door eventually washed ashore about 10 miles west of the calculated splashdown location, consistent with current drifts that would have caused re-surfacing debris to drift Westwards from this location.

Where the specific conditions of a potential Last Known Point and the assumption of lost aircraft power apply, it is possible to create calculated crash maps that permit a much more tightly focused, localized search to proceed. This can result in a much more efficient use of search resources and a significantly higher coverage/probability of detection within this reduced search area.

Of particular interest is that the 'crash map' approach permits the much greater use of ground SAR resources in the search for missing aircraft, as the highly-targeted potential impact sites, particularly for heights of land, can often be examined more closely by mechanized ground SAR teams. Increased cooperation between fixed-wing, rotary and ground SAR resources should provide significant benefits through this combined effort within a highly localized 'crash map' search area.

More information on the preparation of aircraft crash maps, as well as a service to provide these as digital maps, can be found at

http://sartechnology.ca/sartechnology/ST_AircraftCrashMap.htm

Martin Colwell.
SAR Technology Inc.
sartechnology@telus.net
http://sartechnology.ca
*********
Note that Martin Colwell sells the software and services he discusses in his post.

Essentially he's doing NTAP type work without using radar data, by making assumptions about the flight path of the aircraft.

Mike

[InternetSAR.org]

First let me introduce myself.  I am Ken Barbalace the founder of InternetSAR.org (admins can confirm my email address). I joined so that I could answer some of the questions sardak had posed in this thread and to answer any questions others may have.

Looks like the google earth effort to find Fosset has relaunched and started a second search in Canada.

We are using Google Earth, as our imagery application but we are totally independent of Google.

I wonder who is paying for the imagery for the 2nd search.

The imagery we are using to search for Steve Fossett had originally been acquired as part of the original Amazon MTurk search effort but was never used as part of that effort.  We were able to get permission to use it.

The imagery we are using for the Ron Boychuk search effort was donated (I think) by Digital Globe to the Ron Boychuk search team and they asked us to host their imagery search effort.

Currently all costs related to hosting the imagery on the Internet is being paid for out of pocket by myself and the individual who has been helping me process the raw imagery.

Also, who is evaluating the leads now.

Short answer:
We are.

Long answer (this is based on my reply to an email one of your comrades sent me):
We all know about the "I'm SURE this is it" reports to the CAP and anyone else who would listen during the Steve Fossett search effort.  I even tried to bat down a lot of these reports in forums and blogs I was participating in during the Steve Fossett search.  I was especially frustrated by really obvious false hits that would get picked up by the media who was all too eager to get the scoop.  To me it was cases of eyes seeing what minds wanted to find.

At the same time, for many users the biggest frustration was the total lack of feedback, and the feeling that reports were getting ignored.  To address both sides of this issue I built a reporting system that allows the reporting user to track their reports and gives search teams a fully managed reporting channel.

When a report is submitted to us, it is reviewed and scored by seven different individuals from our report evaluation team.  This team consists of a core group of individuals from the original Fossett search who have proven themselves to have good eyes and sound logic.

Over the past three months we have gotten to know each other pretty well and know each other's strengths and weaknesses based on daily discussions of the Steve Fossett search effort on the Google Earth Community BBS.  It has been a three month collective effort to educate the group by discussing reported imagery (around 6,000 posts). In some cases one member of the group would go out get actual photos of contentious objects so that we could see what things look like from the ground (e.g. geologic formations, wrecked cars, vegetation, etc.).  In short we saw a need to educate ourselves and collectively tried to.

The members of the initial level of report evaluation are not imagery analysis experts, but they are very good at weeding out false reports, at the same time they are cautious enough not to totally discount something they can't identify.

Any report that scores high enough on this initial level of evaluation is passed on to a second round of evaluation by a professional imagery analyst who works for the U.S. Military and volunteers her spare time to our effort (apparently looking for downed planes is much more satisfying than picking targets to bomb).  It is only after her blessing is given that a report is passed on to the search team.  Eventually we hope to have a team of professional imagery analysts at the secondary review level, but since we are working with volunteers we must start somewhere and recruit more as we go.

Everyone who evaluates and scores a report also provides feedback on the report in regards to what they think it is and why.  These evaluations can be read by all of the evaluators, the search team representatives AND the user who submitted the report.  Transparency is a key part of the process.  The person submitting the report can follow the entire evaluation process for their report.  This will help members learn what it they are looking at, and give them the confidence to know that their report was fully evaluated.  Basically this gives members an incentive to stay within proper reporting channel.

My estimation is that out of any 100,000 images reviewed, we may get around 100-200 reports depending on the nature of the imagery.  From those 100-200 reports 90%-95% will be ruled out by the initial round of evaluators.  Of the reports that get past the initial evaluation round less than half should (maybe 10% or less) will be passed on to the search team by our imagery analysts.  Of course during times of high publicity and lots of new "recruits" our initial false positive rate will go up, but our two evaluation levels will still weed these out before they get to SAR teams.

As the final step in the process, search team representatives have complete access to the reports for their search effort.  They can, if they want, look at reports at any step in the process and they are encouraged to file their own evaluation or final determination for any report that clears our evaluation process and is passed on to them.  Feedback from SAR teams is important, especially for false positives, because this will help us reduce more false reports in the future.  Feedback is the key to the entire process.

One of the big concerns that got discussed frequently with the Steve Fossett search was maintaining the integrity of the crash site.  The problem was that all of the reported hits from the MTurk process were publicly available. To address this issue, the only people who can see any given report and track its progress is the individual who filed the report our team of evaluators and the search team itself.

I believe that it was the report evaluation part of the process where the Steve Fossett search broke down.  Reports were not getting evaluated and nobody was providing feedback.  Out of desperation some users started bypassing proper channels and reporting their reports to anyone who would listen (willingly or not).


OTHER ASPECTS OF OUR PROJECT

-TRAINING
A volunteer fire department welcomes anyone to join, but doesn't send them into a fire without training.  The same should be case with a volunteer imagery analysis organization. 

Our IA expert is also working to develop training aids, tutorials and tests so that basic observers can test and hone their skills.  We also want to give people the opportunity for advancement by having online tests that allow us to objectively evaluate who should be promoted to the level of report evaluator and senior evaluator.  We figure that for every 100 members we will need five report evaluators.

-BETTER METHODOLOGY
One aspect to our methodology which will eliminate a lot of the bad leads is the fact that our members are able to compare old and new imagery directly in Google Earth.  The big problem with the MTurk process was that for the most part people could not rule out objects by seeing if it existed in the old photo.  I would wager that upwards of 95% of the MTurk hits that were reported could have been eliminated by simply comparing them against old imagery.  Even in instances where novice members are not able to see their object in old imagery, all of our report evaluators have the practice and skill to locate where objects of interest should be in old imagery even when the new and old imagery do not line up perfectly (which is a frequent problem with satellite imagery).

-BETTER IMAGERY
The quality of the imagery is critical to the success of a search effort.  As such we hope to build relationships with the imagery acquisition companies so that we can get better quality imagery.  We may also develop custom software to better process aerial imagery to resolve many of the issues that we frequently have with it (e.g. too dark, distorted, etc).

-BETTER CHANNELS FOR DISCUSSION
Being you all participate in this forum you know how important forums like this can be.  One problem with the original MTurk search effort was that there wasn't a well publicized place to discuss the search effort.  This left lots of little pockets of scattered discussions with many people missing out on helpful advice.

To address this, we have made the Google Earth Community BBS our official form and set up threads there for each search mission we undertake.  We then link to that forum thread from the webpage we set up for each search mission.  This gives people a place to post images they have questions about.  It is also a place to discuss and share information and rumors (hopefully help sort out rumors from facts) or simply to vent frustration. 

-LOW OR NO COST
A very important part of our organization is that it is going to be founded as a nonprofit organization and we hope to be able to fund our operational costs with grants and donations so that we don't need to charge search teams for our services.  InternetSAR.org will be a volunteer service organization, not a for profit company.  Everything we do is intended to keep all costs as low as possible.  We eventually hope to extend this to the acquisition of aerial imagery.

- READY TO ACTIVATE QUICKLY
The key to any search effort is being able to respond quickly. As our systems get built we want to be able to respond to requests for assistance very quickly.  We are building tools to help automate the processing of imagery and shorten the turn around time from when the request goes out until when we can start active searches.  This includes pushing imagery from imagery suppliers to our servers via the Internet instead of waiting for DVDs to arrive via FedEx as frequently happened during the MTurk effort.

- MAKING SEARCH MORE EFFECTIVE
We aren't here to replace traditional search and rescue rather to be another tool in the toolbox for SAR.  The problem with traditional SAR efforts is the logistics of managing and caring for manpower.  If we can keep people at their homes yet allow them to assist in there efforts via the Internet, suddenly the logistical constraints go away and the amount of manpower brought to bare on a search effort can increase exponentially.   In addition we are not limited to daylight hours and good weather conditions. 

SUMMARY
In summary, we addressed the key issues that caused the Amazon MTurk search for Steve Fossett to fail in the following ways:

We are reduced the noise factor of false reports we built a complete report management system and provide evaluators to evaluate reports so that only the very best leads get reported to SAR teams. We also provide a complete feedback loop so that members can see the final disposition of their report.  We made the reporting process as easy as we could and repeatedly drive home the point that reports need to go through our reporting system and must not be reported directly to SAR teams.

To protect the integrity crash sites we limit access to reports to only the individual who filed the report, our report evaluators and the SAR team.

To reduce false positives we are working on developing tutorials and online training to educate volunteers.  Eventually this may include the requirement to go through online training before being allowed to participate in an active search.

We are recruiting SAR and IA professionals to help guide and oversee our efforts.  Eventually this may even include a paid professional staff.

Why am I doing this
I am doing this because I believe that this could put the Internet to use for a higher purpose.  Why should only rich families, and high profile searches have access to massive search efforts?  What about the Ron Boychuks of the world?  What about poorer countries that don't have great search organizations like the CAP? 

I believe that if it is done right and developed properly, an organization can be built up around the Internet that can become a valuable tool for search and rescue efforts around the world regardless of borders. In time it will be an organization that can quickly mobilize thousands of trained volunteers the world over to assist in search efforts by analyzing massive amounts of imagery very quickly and for very little money (regardless of who pays).

The Steve Fossett search proved that the Internet can be used to mobilize thousands of people for one purpose.  Now we will prove that we can organize and train them as well as manage the flow of leads such that search teams can effectively use the results.

[InternetSAR.org]

One constant point of confusion for people and the media is what roll Google Earth (and thus Google) plays in Internet imagery review efforts.  I am sure there will be lots of confusion among CAP members about this as well.

Google Earth is simply a software application like Microsoft Word or Internet Explorer.  It just happens to enable people to look at aerial imagery of the earth along with other geospacial data (like a simplified GIS application).  All default imagery that Google Earth provides is old and was normally taken several years earlier.  Thus individuals using Google Earth by itself outside of a structured effort like InternetSAR.org are looking at old data that would not contain the object being searched for. 

InternetSAR.org uses Google Earth as our primary software application to deliver our updated aerial or satellite imagery to our members via something called a KML overlay file (think of it like a Word document or web page). This overlay file contains instructions for Google Earth as to what imagery to request from our web server and how to position those images on top of Google Earth's old imagery.  This allows our members to compare our new imagery against Google Earth's old imagery.

What is important about this is that in order to review updated imagery for a search mission we are working on, individuals must be members of our site. 

Our members are given very strict instructions (on our main page, on the page for each search mission, on our getting started page and within our Google Earth overlays) that they are to only report findings from within our imagery via our own reporting tools. This allows us to manage and filter reports for search teams via our team of report evaluators and our military trained imagery analyst.

Our members are very explicitly instructed that they are NOT to be reporting their findings from our imagery to search teams directly. Any member who does bypass our reporting channel and takes their report directly to the search team in question is subject to having their account on our website terminated and they will not be allowed to continue participating in our efforts. 

We are very, very serious about managing the flow of information gathered by our efforts and about not generating lots of white noise for search teams.  As such if any search team ever gets an unfiltered report directly from someone participating in our efforts, and the search team provides me with the individual's name and email address we will terminate their account on our website if we can match the information up with an account on our website.  We will also be taking every effort through our training and online tutorials that are in development to ensure our members fully understand the importance of using proper reporting channels.

If you have any questions, concerns or suggestions, about this or anything else related to InternetSAR.org please post them to this thread or contact me directly via PM, our website contact form or call me at 207-797-8202 (I'm on Eastern Time).