In OpenCV, we can use a xml file to describe haar-like features of a specific object. I wanna use this idea to detect palm and fist, now I hava my own xml feature file, and it works well in C and Python(with OpenCV), now I need to move this idea to flash.
Luckily, I found an project named Marilena, it successfully moved the haar-trainning part of OpenCV to actionscript, and this article optimizing flash based face detection optimized Marilena, he turned the xml file into actionscript class itself to speedup the calculation. However, since the Haar data is being hardcoded into the class now it will not be able to use for tracking anything else but faces. I try to read the source code for detail, and process my own xml file into actionscript class file in his way, but finally find something different.
The xml file for face detection is like :
Every tree has only one node "root node", but in my xml file, some trees have two nodes, and the "left_val" above changes to "left_node" sometime, I don't know how to process these nodes. I think I should goto OpenCV C source files for detail to find how to correctly process xml feature file, but where to start, some advices?