by pageauc » Mon Aug 25, 2014 9:46 amLatest project updateHere is project update 4 plus project playlisthttps://www.youtube.com/watch?v=Lc0isSkMcZAhttps://www.youtube.com/playlist?list=P ... DPQ3N1BIFBhttp://youtu.be/P-WpWenJISsYouTube video of progress on my OpenCV Autonomous Robot Project with python coding using the openelectrons pan/tilt RPI camera mount. Currently still working on various behaviors.So far the logic will be 1. start in motion detect mode and look for motion. (might have robot turn once in a while to check other areas)2. When motion found do a local pan/tilt to look for face using opencv face detect.3. If face not found in local area then expand to full range pan/tilt search.4. If face not found the robot turns to look behind (might do this in overlapping scans) 5. If no face found after a specified timeout then robot returns to starting position and goes into motion detect mode again. 6 If face found then lock onto face, track range and position. Move to specified range. Move back or forward to maintain range. Rotate to follow left right movement Etc. 7 if face lowers below a specified angle then robot will approach person (like a dog) to a specified range.8.If face position angle increases beyond specified angle the robot backs up to a specified range from person9. There are other behaviors that will be possible but one step at a time. 10 if face cannot be found after a specified time then goto step 1.You Tube video showing some progress on project. This uses OpenCV, picamera and pipan python modules. No V4L2 interface is used to stream, Only picamera. Currently video resolution is 320x240 so performance is a little laggy so might try 160x120 to see if this improves situation. Lighting is a factor as well as face position. Might need to add OpenCV face profile haas check as well.Note code snapshot has been posted on openelectrons forum under Software.