Analysis of Sonar data
Reply to SaturnV #4296 & cessnapuppy #4297
The time and effort needed to analyse the sonar data is an interesting point as there should be a vast amount of data. I presume that the raw sonar data is first processed by computer to produce a 2D (or 3D?) image of the ocean floor, which then has to be carefully inspected by a human in order to identify a possibly significant shape. I would guess that the scale used on such an image would need to be about 1cm to 2.5m in order for a distinctive shape such as an engine cowling, section of fuselage etc. to be recognised. In which case, a 30cm by 20cm image would represent about 0.004 sq km of the ocean floor; a sonar search over a radius of 75km would require 4.5 million such images to provide coverage of the area with a view in just one direction. Obviously a sonar scan of the front of the engine is going to stand a better chance of being recognised than one from the back. Assuming a human could process 200 images per day, and could continue to carry out this task every day for a month or so this would require 50 man years of effort or more!! Perhaps a team of 200 spending a day per week for 2 years.
An alternative approach was used on a similar data analysis task for an internet project called 'Galaxy Zoo'. The task was to analyse images of galaxies ( a million of them). The instigators of the project set up a web site where interested members of the public could download a set of images, analyse them and send the results back. They were amazed at the response they received; more than 50 million classifications were received by the project during its first year, from almost 150,000 people. The loss of AF447 has generated a lot of public interest, perhaps the 'Galaxy Zoo' approach could work here on the sonar data and complete the analysis in months rather than years? Could the Pprune network be a good starting point to get the ball rolling?