Analyzing Videos for Fun and Profit

Although these panes do give you lots of raw data, there really isn't any analysis being done yet. Clicking the View→Data Tool menu item pops open a new window where you can use this raw data to figure out what is happening in the video.

Figure 5. The Data Tool window provides a set of analysis tools to look at the raw data pulled from the video.

The Measure and Analyze buttons at the top of the plot window provide a set of different sections that you can display. This gives you several different analysis options for your data. Clicking the Measure button gives you the options Coordinates, Slope and Area. When you select one or more of those options, values are displayed at the bottom of the plot representing each of the selected options.

When you click the Analyze button, you get a drop-down with the Statistics, Curve Fits and Fourier Spectrum options. Each of those options actually opens a new pane in the Data Tool window. The exception is the Fourier Spectrum option. This option actually opens a new window where a Fourier analysis is done on the raw data from the tracked object.

Figure 6. Tracker also can do a Fourier analysis of the observed motion.

The Statistics option gives the data displayed in the top right pane. This is the set of descriptive statistics—that is, the maximum, minimum, mean and standard deviation, among other values.

Selecting the Curve Fits option gives you the pane at the bottom of the plot window. Here you can select what type of curve is being fitted to your raw data. You can use a number of default functions to try to fit your data. In this case, the linear function is the obvious choice for a fitting function. You can see the function being used, x=A*t+B in this case, along with the calculated values for the parameters A and B. If you are tracking an object that has a really odd motion, you can click the Fit Builder button to pop up a new window where you define a new fitting function.

Figure 7. You define your own fitting functions as part of the data analysis.

What do you do as part of a new project? Let's say you want to analyze the pendulum video from the sample video folder. You can set some calibration items within the video by either selecting the relevant buttons at the top of the window or selecting the items under the menu entry Track→New→Calibration Tools. You can set up items like a calibration stick or a reference set of axes.

Figure 8. You can set up calibration tools within your video to be analyzed.

One way to do the analysis is to step through the video, frame by frame, selecting the object to be tracked within each frame. This may be the only way to collect the raw data, but humans are lazy. If the object being tracked is relatively consistent across the duration of the video clip in question, you can try using the autotracker. Clicking the autotracker button will pop up a new window where you can control what is being tracked.

Figure 9. The autotracker tries to follow an object from frame to frame automatically.

Tracker helpfully displays reminder information on how to use the software. In this case, it reminds you that you need to press Ctrl, Shift and left-click the mouse button on the object of interest. If you then click the search button within the autotracker window, it will move through each frame, looking for the object in question. Once it is done, you should have a full set of data from the tracked object. From the plot in Figure 10, you can see right away that the pendulum follows the expected motion. You also can track multiple objects by creating a new track for each object.

Figure 10. The autotracker can generate all of your raw data automatically.

Now that you've taken a quick look at Tracker and the kind of analysis you can do with it, you should be comfortable enough to start experimenting with it. Don't forget to share any interesting ideas you come up with on the community page and add to a great resource for citizen science.


Joey Bernard has a background in both physics and computer science. This serves him well in his day job as a computational research consultant at the University of New Brunswick. He also teaches computational physics and parallel programming.