If you still spend your Friday nights staring at an untuned TV in hopes of spotting a signal from an alien world, you haven’t yet discovered the Stardust@Home project. For the past year and a half volunteers have been sifting through microscopic pictures of tiny portions of aerogel plates retrieved from the Stardust spacecraft trying to find evidence of collisions with interstellar dust.
Admittedly, I have spent a few hours doing this, and being a naturally lazy guy, I immediately started pondering ways to automate the search. Apparently software to do just that was part of the original Stardust proposal, but in the end they didn’t get the budget to develop it. They also claim the task is not susceptible to programmatic distillation… a plausible excuse.
While a completely automated searching program would be a daunting task, tools to make it easier for humans to pick out the desired anomalous features seems quite a bit easier. The website teaches us that one of the primary features of a collision track is that it extends below of the surface of the gel, so my first thought was to stack the frames coloured by depth. This made the tracks easy to spot within the sets of images I chose. (If you haven’t seen Stardust@Home movies, these will seems quite unremarkable)
-
- Composite of a Stardust@Home calibration movie (unknown id)
-
- Composite of Stardust@Home calibration movie 7575333V1
-
- Composite of Stardust@Home calibration movie 9914558v1
(These were done manually, tediously, in PhotoShop. Frames were layered and to each the Find Edges filter was applied, then all blue and green removed leaving red outlines. The hue of each frame was then shifted by an ever increasing amount, giving each frame a different colour. Finally, the blending option on all frames was set to Lighten.)
Granted this is just a small sample. The movies I chose may very well be ideal candidates for this type of filtering. Regardless, it gives me hope that a filter of this sort would be useful in helping spot tracks. The next step would be to write a little app to do this in real-time to see how this technique holds up.
Regardless, it’d be nice to see something like this and other filtering options alongside the current focus movies. Here’s a list of other filtering ideas. Many of these could be combined.
1. Layer the frames and allow interactive dragging of the top frame with decreasing movement of the lower frames to produce a pseudo-3D or parallax effect. This might help the eye pick out features with depth.
2. Side view. Take the same row of each frame and stack them to produce a new side-on image, and allow for scrolling through the rows. Again, this may make it easier to spot details below the surface, especially if layered.
3. Allow the viewer to select the frame, and using the known optical properties of the microscope, calculate the expected blur on higher and lower frames and subtract it away. Perhaps this is getting a little too fancy?
4. Correct Tilt. I find sloped movies the most difficult to search, so by specifying a location that is in-focus on two different frames, the entire set of images could be tilt-corrected, making it easier to spot tracks.
5. Zoom, brightness, contrast controls.
.. til next time.