Using Hardware Acceleration To Improve The Efficiency Of Face Detection

With the large-scale use of intelligent security monitoring, the intelligent security monitoring system which integrates the Internet of Things, cloud computing and other new technologies comes into being. Intelligent security monitoring system provides intelligent,secure, reliable and convenient service for banks, schools, communities, transportation and other places. As a security monitoring system intelligent "eyes", security surveillance cameras develop to be more comprehensive, intelligent. More IP(Internet protocol)cameras with lower stream are also developed. Now many security monitoring systems are equipped with advanced machine vision technology and image processing technology,including face detection and face recognition algorithms. It can improve the monitoring capability efficiently, especially in terms of information filtering, such as security incidents in real-time alarm and video surveys. In this paper, in order to improve the efficiency of face detection, it studies the efficiency of face detection based on face detection algorithm which used in ARM- embedded security surveillance cameras. The detailed results are as follows:First, it studies the factors affecting the face detection efficiency in detail, including two factors: the hardware factor and the software factor. The performance of CPU as the main hardware factor affects the efficiency. Software main factors are the human face detection algorithm efficiency at the PC side and the ARM side. Meanwhile the running time of face detection algorithm at the PC-side and at the ARM side are both calculated.Finally, it is found that hardware acceleration by a "mask" approach is more effective than software runing on improving face detection efficiency.Second, the effect of using image preprocessing on face detection efficiency is studied.In this paper, using equalizehist function as the method of image preprocessing, the process of the function is as follows: 1) analysis of the brightness changes of the image; 2) the image brightness correction; 3) transforming the gray-scale of image. It is found that the use of the image preprocessing can greatly improve the efficiency and accuracy of face detection on the PC side and the ARM side.Third, it does the research about the impact of hardware acceleration on improving face detection efficiency.. Hardware acceleration is that using existing hardware processor chip is instead of software modules to achieve a specific algorithm, thereby improving the operation efficiency of image processing. Here SOC chip used. The SOC chip is equipped with the image preprocessing algorithm, the function of equalizehist. By function Get TickCount(), we can obtain the running time of the image processing. It is achieved that image processing time can be reduced to 0.4384 ms at the ARM side by hardware acceleration while it usually use13.1628 ms to running the image program at the ARM side..