An embedded system with realtime surveillance application

To reduce the manpower and response time for surveillance systems at low cost, in this paper, an ARM-based embedded system dedicated for unattended realtime moving target detection is constructed. The comprehensive procedures in building up an embedded system such as setup environment for cross-compilation, migration of Bootloader, migration of Linux-2.6 kernel, fabrication and migration of root document system and setup of peripheral driving devices have been presented. The algorithm of image background subtraction for moving target detection and tracking technology has been presented. In this embedded system, for two consecutive 640×480 image frames captured by camera, if the difference of 8-bit gray level value of a pixel at the same position is greater than 32, that pixel is marked as a “moving pixel”. When there are more than 500 moving pixels in two consecutive image frames, the camera is triggered to take pictures because the system assumes that a moving “invader” appears. Consequently, the system will transfer the taken invader´s pictures to “The Cloud” through WiFi to prevent the pictures being destroyed by the invader. The constructed embedded system can be used in a security system and other applications with proper modifications.