| Main > Lightguns |
| DIY projects with other than IR led tracking |
| << < (2/2) |
| Howard_Casto:
--- Quote from: JayBee on August 11, 2020, 02:48:46 am ---Very nice tests you did there. Yeah the amount of data to process and all the issues that can come from this solution aren't easy to overcome I guess. --- Quote from: Howard_Casto on August 10, 2020, 03:57:18 pm ---If one is using a camera, IR filtering is probably the best way to go. Basically it filters off a buttload of image data before the camera even starts processing, which increases responsiveness. We were discussing trying IR lasers and a stationary camera last year but we had trouble sourcing a ir laser powerful enough that wouldn't destroy our retinas. --- End quote --- But would IR filtering be really useful here without any IR emitting source? I also tried the laser stuff a bit, and true, it's not working very well, especially with non glossy screen that don't reflect the laser enough. --- End quote --- I'm not exactly sure what you are asking. Blocking out everything but IR light essentially makes it a black and white camera that seldom has any white in the image but the targeted tracking points. So yeah, on paper at least that really makes it easier to process than, let's say a full color image where a very specific color is the tracking color. It's just less stuff to parse. sniiki basically asked why IR was being used.... well amongst other reasons that's why. |
| JayBee:
--- Quote from: Howard_Casto on August 11, 2020, 01:57:11 pm --- --- Quote from: JayBee on August 11, 2020, 02:48:46 am ---Very nice tests you did there. Yeah the amount of data to process and all the issues that can come from this solution aren't easy to overcome I guess. --- Quote from: Howard_Casto on August 10, 2020, 03:57:18 pm ---If one is using a camera, IR filtering is probably the best way to go. Basically it filters off a buttload of image data before the camera even starts processing, which increases responsiveness. We were discussing trying IR lasers and a stationary camera last year but we had trouble sourcing a ir laser powerful enough that wouldn't destroy our retinas. --- End quote --- But would IR filtering be really useful here without any IR emitting source? I also tried the laser stuff a bit, and true, it's not working very well, especially with non glossy screen that don't reflect the laser enough. --- End quote --- I'm not exactly sure what you are asking. Blocking out everything but IR light essentially makes it a black and white camera that seldom has any white in the image but the targeted tracking points. So yeah, on paper at least that really makes it easier to process than, let's say a full color image where a very specific color is the tracking color. It's just less stuff to parse. sniiki basically asked why IR was being used.... well amongst other reasons that's why. --- End quote --- Just quoting what Sniiki was saying: [...]using something else than IR leds to track the monitor? He wasn't talking about IR solution at all, but instead screen tracking solution. Your suggestion of adding a filter only works with IR tracking points. Not in this case. |
| lightgungamer:
Hey. That looks really good! I’m not sure that using a banana for a gun will catch on but what you have on the screen looks excellent. I would guess that the lag could be reduced but how consistent is it at finding the screen accurately? Would it work well on different monitors or does it need to be coded for the one you are using? |
| caiom:
One of the biggest challenge in this case is the latency - the time from when you hit the trigger to the time the mouse position is received. It is really hard to find a USB camera with low latency, I don't think there is one easily accessible. After that you still have the update rate (camera FPS), image processing time and so on... Your processing time is 15 ms, which is nothing compared to the total latency (~1000 ms). The Sinden processing time is 7 ms, but I guess the Sinden's total latency should be in the range 50-100 ms. So I don't expect good latency from Sinden. I also thought of using a IMU (accel, gyro) to estimate the moviment, so that the camera gives an absolute position and the IMU gives a relative one. This is analogous to a car positioning system (GPS gives the absolute position). I guess it is not hard to find a 1000hz IMU. In this cenario, due to the image grabbing delay, we would never use the actual image information to update the current position, rather we will use it to correct a past position and propagate this correction to the current positon using past reading of the IMU sensor. What do you guys think? One, question, how bad the IR based solutions latency? Best regards, |
| JayBee:
--- Quote from: caiom on September 23, 2020, 01:34:09 am ---One of the biggest challenge in this case is the latency - the time from when you hit the trigger to the time the mouse position is received. It is really hard to find a USB camera with low latency, I don't think there is one easily accessible. After that you still have the update rate (camera FPS), image processing time and so on... Your processing time is 15 ms, which is nothing compared to the total latency (~1000 ms). The Sinden processing time is 7 ms, but I guess the Sinden's total latency should be in the range 50-100 ms. So I don't expect good latency from Sinden. I also thought of using a IMU (accel, gyro) to estimate the moviment, so that the camera gives an absolute position and the IMU gives a relative one. This is analogous to a car positioning system (GPS gives the absolute position). I guess it is not hard to find a 1000hz IMU. In this cenario, due to the image grabbing delay, we would never use the actual image information to update the current position, rather we will use it to correct a past position and propagate this correction to the current positon using past reading of the IMU sensor. What do you guys think? One, question, how bad the IR based solutions latency? Best regards, --- End quote --- Not sure about the other solutions, but my IR solution has a total processing time of around 7ms thanks to the very fast IR cam we use and the calculation I do ;) I still have to find way to measure the total latency accurately, but from my tests it's under 100ms. One of the few advantages IR/camera based solutions have compared to common crt tech is that the position is refreshed all the time from the cam, so whenever a button is pressed both button press and position are sent right away. With crt tech once the button is pushed, you need to wait for next full frame to become white and check when the tv beam hits the sensor. So depending of the tech, it can take as long as 3 frames (not including the game and controller latency), and is harder to handle because it relies on screen resolution and timing. You're right about the normal cam solution, it's quite hard to have a good enough camera, and be able to process as many frames at it would need to get good latency. It's funny you're talking about IMU, I'm actually trying to add it to my solution too, to make detection faster and more accurate. |
| Navigation |
| Message Index |
| Previous page |