Abstract:

Face detection is an important part in intelligent video surveillance systems. The Adaboost face detection algorithm of OpenCV can process videos in a realtime way. Because of the languages and platforms, very few systems are compatible with such a function. A module of face detection that supports multiple program languages is presented,and the P/Invoke techniques in the .NET and the improved JNI methods for calling the OpenCV face detection module are explained in detail. This module solves the problems of parameter passing among .NET, Java and OpenCV for color images and can be simply integrated into most existing videobased application systems. It also defines a standard interface for module scheduling and can be referenced by more crossplatform applications. The experimental results show that the video systems which integrate this module obtain a high detection rate and realtime processing speed with limited system resources.

Key words: native interface;OpenCV face detection;video surveillance;performance evaluation