Based on the need of automatic observation of human behaviour, methods are discussed which are able to measure human behaviour and which are able to control this behaviour. The method “Using computer vision to measure human behaviour” which has the best prospects of success for measuring the human behaviour is analyzed in detail. Based on these analyses an implementation is provided which allows measuring of human behaviour based on low rate video streams in real-time with an accuracy of about 80%.
This framework provides a simple, configurable and extensible process chain mechanismn to analyse video stream data. Up to now, filters and analyse steps are implemented that are needed for the Behavioral Rating of Dancing Human Crowds based on Motion Patterns. The framework is extensible be additional filters and analysis steps. It supports different kind of video sources like web cam, images, wmv videos and so on . Is based on OpenCV, Emgu and the Sebarf Library from remoteobjectslinq.codeplex.com/
. It is developped in C#.
For further information see either the general documentation http://emotiondetection.codeplex.com/Release/ProjectReleases.aspx?ReleaseId=28871#DownloadId=71932
or the writen paper http://emotiondetection.codeplex.com/Release/ProjectReleases.aspx?ReleaseId=28871#DownloadId=74550UIEasy to use
Configuration of a process chain
new AlgorithmViewer("MotionByDiffColorsBySocialHeuristic.xaml", "MotionByDiffColors").ShowDialog();
by XAMLProject Team