Abstract:
The giant planet Jupiter is the telescope-user’s delight with interesting cloud belts,
cyclones, anti-cyclones and other fascinating features. This paper describes an
inexpensive and accurate automatic technique developed using image processing
methods to detect the variation of features on the planet Jupiter. The study was focused on
the two main features namely the Great Red Spot (GRS) and shadows cast by Jupiter
moons on the planetary disk. GRS is a very fast storm in Jupiter’s atmosphere rotating anticlockwise
and completing a full rotation in six days. Observing the GRS is one of special
treats in astronomy. A method of detecting the GRS was developed using both image
processing and clustering techniques. With the help of this developed system, one can
detect the GRS well and measure the size of its long axis within an error range of 0.1%.
Using this new technique 35 Jupiter images captured during the period 2010 – 2014 were
analysed. In 2004 the size of the GRS was found to be 16570 km across. It is found that
the GRS has begun to shrink and from 2010 to 2014 the red spot had shrunk at a rate of
375 km per year and its shape is gradually changing from an oval to a circle. These values
are in agreement within an accuracy of 97% with those of other researchers obtained
through manual methods.
Another method was developed to automatically identify Jupiter’s Galilean moons by using
the shadows of moons on the planet. This is important because sometimes even an
experienced astronomer may not be able to identify the exact moon which transits by
looking at the shadow. With the help of this developed automatic method, anyone can
identify which moon belongs the respective shadow cast on the Jupiter disk. Both image
processing and Artificial Neural Network methods were used to develop this method. The
Artificial Neural Network was trained using supervised learning to classify the largest four
moons of the planet Jupiter, Europa, Io, Ganymede and Callisto. The trained network was
capable of identifying the moons at a success rate of 95.5%.
This technique can be further extended to detect other features of Jupiter as well with the
combination of both image processing and Artificial Neural Networks.
