The Research Of Face Detection Based On Z_floatboost In Embedded Platform

This paper explores and realizes the typical image application of human face detection in embedded platform.Embedded platform is usually made with lower electricity, it can not keep heavy work for a long time. Besides, network transmission on embedded platform costs much. Thinking of the differences between the embedded platform and the traditional PC terminal, combined with the principle of Skin detection, the paper raises a way that divide images into smaller areas for further processing by YCbCr model. In this way a great amount of image processing operations was reduced. At the same time, this paper adopts the OpenCV library with high efficiency and low complexity functions. In the code selection, also using a higher efficiency of the C/C++ instead of Java to achieve the related functions.Algorithm design in this paper is based on the improved Haar_Like features and Z_floatboost algorithm, the original Haar_Like features cannot accurately reflect the face's arc characteristics, so the improved Haar_Like features is putted forward. Z_floatboost is proposed to solve the problem of Adaboost's non monotonic and high false detection rate of Floatboost.In the specific implementation, the paper firstly completes the improved algorithm in the PC, and collect the positive and negative samples to train the classifier file for using in the Android platform. Android application development is completed in the Eclipse environment, through the JNI, the application finshed the interaction between Java and C/C++ which is used for image processing.The final results reaches the effect of real-time detection by the opening camera. Compared to Android's method of the face detection based on the eye's size, the method in this paper improves the detection rate, in addition, the false detection rate also has been reduced.