Embedded Video-based Vehicle Tracking System Design

Along with the rapid development of the transportation industry, intelligent transportation system in traffic management is playing an increasingly important role. Video-based vehicle tracking technology has become an important research direction of the intelligent transportation system, also being the world’s research focus. In this paper, embedded technology, video detection and tracking technology are used to design embedded solutions for vehicle tracking system which can be applied to track the moving vehicles.This paper gives the general architecture of the system, including hardware platform, bootloader, embedded Linux system, applications. Including Samsung S3C6410X microprocessor core, a CCD analog video capture card based on the TVP5150A; Linux development environment is built in the PC, the Linux kernel is cut and cross-compiled, then being downloaded to the Nand Flash of the S3C6410X development board; also applications are developed and compiled to execute in the board.The vehicle tracking algorithm is transplanted on embedded systems, video-based vehicle tracking terminal platform is designed for the actual application. Software runs on the Linux operating system, the framework of the vehicle tracking algorithm is given, then programmed in C++language through referencing OpenCV. The foreground of vehicles are extracted with the method of gaussian mixture model background subtraction and morphological gradient texture extraction. Vehicle tracking using the method of Camshift, adding Kalman filter model to predict the target position in the current frame. The video capture program is programmed in accordance with the norms of Linux V4L2, also the LCD display application is programmed, through cross-compiling and porting, then debugging the applications.The vehicle tracking program developed in the PC has been ported to run on the S3C6410X after cross-compiled, the dynamic targets captured through the CCD camera are tracked and displayed on the LCD screen. The Linux performance analysis tool Oprofile is used to get the result of function time-consuming analysis when the program is executing. Then find out the hotspot function of the program, according which the vehicle detection and tracking algorithms can be optimized to improve the video processing speed.