Research And Implementation Of Human Face Detection And Tracking System On Android Platform

With the rapid development of artificial intelligence and computer technology, the research of computer vision has made considerable progress. As the foundation and core technology of the field of computer vision based on human face, face detection and tracking technology are attracting more and more researchers to join in its study, and continuous research results appear.This paper uses the Adaboost face detection algorithm based on the characteristics of the MB-LBP. MB-LBP can effectively expressed a variety of image structure by encoding rectangular regions through LBP, such as edges, lines, spots, flat areas and corners, etc. Compared to the rectangular Haar-like features, the number of MB-LBP feature set is much smaller, so that the weak classifier training speed have increased significantly. The face recognition ability of the MB-LBP features is more significant, thereby reducing the number of classifier layers and weak classifiers of each level, the classifier performance will not decline. The experimental results show that the Adaboost face detection algorithm based on the MB-LBP features in the no loss of detection accuracy, detection speed significantly improved. This paper uses CamShift target tracking algorithm to achieve real-time face tracking based on skin color. The component H of the HSV color model is used to calculate the color histogram of the face region, at the same time, in order to remove the area close to the hue information of the face region, the information of saturation and value are added. The experimental results show that, in the way of adding saturation and value information described in this paper, the robustness of the CamShift face tracking algorithm has been significantly improved.Finally, the face detection and tracking system on Android platform according to the intelligent docking station for Android phones is implemented based on the improved face detection and tracking algorithm and OpenCV libraries. This system fist capture video information through the front camera of the Android phone, then detect and track the human face, and finaly control the docking station rotating by USB peripheral. Due to the need of rewriting the face detection and tracking function code in order to implement the improved face detection and tracking algorithm and making it run in the native to improve the speed, we use OpenCV C++APIs. Because Android applications are written in the Java programming language, that is involved with Java and C++call each other, so using JNI interface to realize this purpose, while the Android NDK compile the native code.