The second trailer for Star Wars VII: The Force Awakens looks exciting. In case you haven’t seen it, you can watch it here. So, how about an analysis? Can I estimate the speed of the Rey’s speeder as it moves across the desert? Let’s try.
The first thing I need is a scale. This is pretty tough. Really, the only thing I can use to set the size of the motion in each frame is the size of the speeder itself. In my previous analysis of the speeder (from the first trailer), I estimated the size of the large part of the speeder to be about 2.8 meters long. From this, I will use a rough value of 4.0 meters for the entire length of the speeder.
After that, it’s just a simple process of shifting the coordinate axis origin to accommodate the panning camera (I used calibration point pairs in Tracker Video Analysis).
Here is a plot of the position of the speeder as a function of time.
You can see some problems with this data. The apparent speed at the beginning and end of the motion look to be different than in the middle. This could be explained by a couple of reasons. First, I think there might be a problem with Tracker’s axis over large changes in frame. It seems like the axis deviates from the horizon towards the end of the video. I’m not sure if this is a problem with my adjustment for the panning of the video or with the software. Second, the changes in motion could be due to changes in scale. Here, let me draw a sketch showing both the motion of the speeder and the camera.
In the middle of the speeder’s motion, the speeder would be slightly closer to the camera giving it a larger apparent size and a larger apparent speed. Ok, so that could be the problem.
As an estimate, let me just determine a value for the speed based on my initial estimate of the scale and the slope of the position graph near the middle (where it looks more constant). This gives a speeder speed of 74.7 m/s (167 mph). Of course, this is just an estimate. If I approximate the speeder size as 4.0 +/- 1.0 meter, this would give a speed of 74.7 +/- 18.7 m/s (167 +/- 41.8 mph) – which is still pretty fast. Here is a quick review of uncertainty calculations if you are interested.
Based on the scale of 4.0 meters, I can also get an estimate of how far the speeder moved during this time. It’s about 858 meters. That’s pretty far. So, how close would the camera have to be to the speeder in order to get a slight distortion in apparent speed as shown? I’ll leave that up to you to estimate as a homework question. Here is the data from the video in a Google spreadsheet (you might need that to complete the homework).
Just for fun, let’s look at one more thing. While correcting for the panning motion of the camera, I also noticed that the camera appears to jump up and down a little bit as though it were manually being moved. I suspect that this whole scene was done with CGI and not a real camera, but I could be wrong.
How do you measure the camera motion? This is pretty easy to do using Tracker Video Analysis. Without setting the scale or the axis, I can just track the motion of some background object (in this case I picked part of the crashed x-wing fighter). It really helps to use Tracker’s autotracker feature.
Here is the horizontal motion of the camera (the distance units are just “pixels”).
There’s not too much to see here. The camera just pans and any “shaking” in the horizontal direction is difficult to see within the panning motion. However, you might be able to get a better look at the motion. First, here is the data from the video. What if you get a baseline motion using the average panning speed. Then you could plot the horizontal deviations from this average. Honestly, I’m not sure what you will find. It’s possible that there is no horizontal shake. Ok, now for the vertical motion. Normally, I would plot the trajectory of background (x vs. y) but in this case I am going to plot vertical position versus time.
But what does this tell us? I’m not sure – but it’s homework for you.
- Create a plot of the vertical jump as a function of time. Does this show anything cool?
- Are the jump sizes random or do they create some pattern?
- Create a Fourier Transform of the data to get an estimate of the frequency of camera shakes. How does this compare to a hand held camera? Just for comparison, here is a video that uses “fake camera shake”.
- If you found any horizontal deviations, how does the horizontal shake compare to the vertical shake?
In the end, you want to answer the question: Is this a CGI (Computer Generated Imagery) or a partially real video? If there is evidence of fake shake, that would point toward CGI.