ANALYSIS & OPINION
Streamlined data analysis finds threats on the runway
16 November 2007Tweet
They may not look like a threat to homeland security, but birds have brought down at least two planes within the past year. The unsuspecting birds were sucked into the plane’s engines when they flew too close to the flight path, causing an RAF Tornado to crash in autumn 2006 and forcing a commercial airliner to land last April.
Like the most dangerous terror suspect, their every movement must be analysed to predict where they will hit next, and this process frequently involves complicated mathematics to create meaningful predictions from large volumes of unadulterated data.
Typically, researchers such as Richard Budgey from the Central Science Laboratory, spend a couple of weeks on location with a mobile radar unit that captures the coordinates of each bird in a flock and stores the information, which can amount to 100Mb a day, in an MS Access database.
However, this data is not very useful to researchers by itself, so Budgey and his team convert the data into a file format, called ‘shapefiles’, that identifies each bird and tracks its movement over time. This can then be interpreted by geographic information systems (GIS) to visualise the mass movement of flocks of birds over a 24-hour time period, which could help the security staff to determine how to direct the birds away from flight paths.
‘It helps us to identify problems that we weren’t even aware were there,' Budgey told scientific-computing.com. ‘We can find out what’s attracting them to live at these points, and the solution can be as straightforward as moving a food source.’
Previously, this data conversion was a complicated process that could last for 24 hours for each day’s worth of data. The data had to be fed through a different program at each step of the process, creating an inefficient workflow and creating potential inaccuracies. ‘It was easy to get lost in the process – there was just too much scope for error,’ Budgey said. ‘We needed to spend a long time setting up each new part of the process. It was a bit of a nightmare really.’
To overcome this problem, the team recently implemented Matlab from The Mathworks, which could perform all the steps of the process in one environment. Matlab reads the database using the Database Toolbox, and then trims and tailors the data, removing noise and eliminating records where the radar had picked up another object by mistake.
The program then calculates the speed and direction of the bird, and converts this into a line that represents its path. The Mapping Toolbox then converts this information into the necessary file format.
Matlab was particularly suited to the manipulation of the very large matrixes that contained the data. In addition, it allowed Budgey to write a graphical user interface that is intuitive and easy to use. Overall, it reduced the analysis time of a day’s activity from 24 hours to 30 minutes.