Config

Important

This information is only relevant for all software versions below version 0.23.0

With version 0.23.0, these settings have been removed, as they are no longer needed.

Here, Motion Detection and Dart Detection are explained, so you can better adjust them to your needs.

Let us start with the two most important ones first. The Diff Image > Threshold both under “Motion Detection” and “Dart Detection” is used to calculate the difference between the image before the dart hit and the image right after. For optimal results this image should be crystal clear. That is, there should be a clear outline of the dart visible but there shouldn’t be any artifacts in the background. You can check that best while throwing some darts and observing the Motion tab in the Board Manager. Higher Threshold will typically result in less visible dart contours but also in less background. Lower Threshold values will result in clearer dart contour but at the expense of picking up background. Set it so that you are right in between.

It is generally recommended to keep the Diff Image > Threshold the same for Motion Detection and Dart Detection, but it can have a positive effect to set them differently."

Play with these values to optimize the diff image. It will have quite a big impact on the detection overall.

Next is the Gaussian Kernel. It is used to blur the diff image to get rid of background noise. You can only set it to odd values. Higher values will take more time to compute, which is why I default them to 3 and 5 for Motion Detection and Dart Detection, respectively. They do not have a big impact on the detection, so you should keep them as is, if you do not face any problems related to it.

The Object Detection holds the config parameters related to the differentiation between dart and hand. Very simplistically spoken, a dart is smaller than a hand, and that is how the software detects it (believe it or not). So the Dart Size: Min and Dart Size: Max are values that define the boundaries of how small an object at least has to be to be detected as a dart as well as how large it is allowed to be at max. These values are representing areas in the image but you can think of them as the total number of pixels in your image that makes up the dart. I typically put the Hand Size: Min to be a lot higher than the Dart Size: Max but there is no real reason for it. Once an object has more pixels than this value, it is considered to be a hand.

In the Dart Detection, there is a sub-section called ROI Detection. The ROI is the region-of-interest in the full-resolution image. I use the motion detection which runs on a much smaller resolution (to be fast) to calculate where the dart would be in the full-resolution image. To figure out the ROI, I combine all the so-called blobs (objects in the image made up of pixels) that exceed a certain size (again in pixels). The ROI is simply the bounding box (a rectangle) of those combined blobs. I typically set this quite low, lower than the Dart Size: Min. Keep in mind that the number of pixels here refer to the motion resolution, which is 4 times smaller than the detection resolution.

Lastly, there is the Hough Line Detection. This sub-section holds all the values around the algorithm that fits the line into the dart barrel. The Threshold is quite important and I have not really figured out how to scale it properly with the chosen resolution. I typically use values in the 50s for my setups (running at 800x600 and 1280x720). If you set this value higher, the algorithm will find fewer lines because only lines with a certain number of votes are considered. The threshold defines how many votes a line needs. The Min Line Length and Max Line Gap should be quite self-explanatory. Again, they refer to values in pixels. So, how long a line at least has to be, to be considered, as well as how far apart two lines are allowed to be, to be combined into one single, long line.

You can now set the motion/scale config parameter in the Board Manager. The default value for that parameter is 3. For all of you guys sporting a Raspberry Pi setup, it is worth while setting this to 4 to increase your main FPS (not the camera FPS). The motion/scale parameter defines at which resolution the motion detection (dart vs. hand) is running. The motion resolution is simply the main cam resolution divided by that scale parameter.

Warning: If you try to change the motion/scale parameter while the board is running, the board will crash. This will be fixed in the next board client version. Until then, please do the following. Stop the board, then set the motion/scale to a different value, then Restart. You should be able to Stop and Restart with the buttons in the Board Manager.

For the lazy ones among you, you can also change it while the board is running, resulting in the crash. The autodarts service (on Linux) should simply restart after the crash with the new parameter, which has the same effect as doing it as I have described above. It’s just a little more crude .

This should boost the FPS significantly for low-end devices. In my tests, I could also set it down to 5, but I had the feeling that 4 runs a bit smoother in the motion detection. 3 and 4 had no difference at all in my tests.