openframeworks - Detect multiple bodies in Kinect? -


i working kinect in openframework using ofxkinect addon, great , plenty fun!

anyway looking pointers or direction when dealing multiple bodies on screen. thinking of making rect around each detected body , when rects intersect happen, effect or anything.

so looking ideas or point me right direction of detecting multiple bodies when using kinect.

right based on depth image kinect go through each pixel , create bunch of smaller rectangles padding , group them in larger rectangle bound if separate rectangle group. not ideal deals pixel values , not seperating bodies eachother , not giving me results looking for.

so ideas appreciated!

if want use ofxkinect quick solution threshold on depth , assume bodies , no other objects within depth range. should make easy use opencv's contour finder isolate outlines of bodies , bounding rectangles. if rectangles intersect(and ofrectangle math you), trigger reaction need. don't forget once if effect isn't showing already, otherwise trigger effect multiple times per second while 2 bodies' bounding rectangles intersect.

you try bit more hardcore , using ofxcv(not ofxopencv) tap hog functionality. slow in , not ideal depth map, can run in every few seconds detect person , depth, keep tracking movement.

personally, if want track people kinect recommend using ofxopenni if provides scene segmentation feature , if don't track skeletons can still useful information pixels pertaining each body , they're centre of mass. i'm guessing microsoft kinectsdk has similar feature , there should of addon, it's windows only.

ofxkinect/libfreenect not offer people detection features, need roll own.


Comments

Popular posts from this blog

.htaccess - First slash is removed after domain when entering a webpage in the browser -

Automatically create pages in phpfox -

c# - Farseer ContactListener is not working -