|
Project Information
Featured
Downloads
Links
|
Autonomous Sentry Gun ProjectWelcome to the project(03/05/10) Version 3.1a is up!The software is now universal for the following protocols: - Pololu 16channel USB - MiniSSC - Micro Maestro We're hoping to have Arduino and FPGA controllers integrated shortly. If you're interested in finding out more, please head to our forum at: http://members.upc.nl/a.kutsenko/feedback.htm For images and video head to http://members.upc.nl/a.kutsenko/pictures.htm - The Weaponsgrade Kavelaars. Project WishlistIf you're looking for a place to start, here are a few things we're aiming to integrate soon. 1 - HUD Information Once the above camera/calibration panel combination is in effect, we'd love to be able to drop overlays over the screen showing more information. This function is already built into the software with andrew kirillov's motion detector, just a question of intergrating it into the main form. Small project but would be really nice to have. Things that could be shown are:
Note: The servo's usually have 2 to 3 times the step potential than the camera can see using a standard resolution webcam. 2 - Stereo Vision Distance Calculation Using stereo vision we will be able to calculate distances and develop projectile prediction for dispensers that use slower ballistic projectiles. This will make our software one of the most powerful open targeting applications available. 3 - Object Targeting We're always looking to make our targeting smarter. We can find moving objects, we'd love to be able to recognise the same object in sequential frames. As USB webcams usually operate at 30fps, we could use a path history and estimate the direction and speed (of the center of the object at pixels per second). The aim of this is to make it somewhat possible to estimate the next position and eventually ensure target and slow projectiles meet. By getting the Average Colour of a moving object, if it stops moving, the software will still know it's there. 4 - Object Recognition Currently the project is set to target the biggest cluster of pixels that the software recognizes as changing colour. Integrating object recognition into the software will give more accurate or smarter software. Breaking this into stages we could get: Level 1 - Basic differentiation. Knowing the difference between a human, car, truck, animal etc. Level 2 - Advanced differentiation. is it Adam or John, is it a Mazda or a Bentley, dog/cat/cow etc. 5 - Manual Controls Some users have expressed the desire to be able to control their turrets manually in certain situations (eg to target a particular static object) 6 - Semi Automatic Dispenser Control So that dispensers with mechanic firing mechanisms can be used via a 3rd servo. That’s it for now, but if you’ve got any other ideas or requests we’d love to hear them, and add them to the list. |