FPGA implementation of a real-time image tracking system

摘要

This article focuses on realizing an object tracking and image capture system on a FPGA. The system processes each picture with size of 640×480 pixels which is coming from input image data captured by an external CCD camera. Firstly, the RGB format of input image data is transformed to HSV format through color space transformation. Secondly, the value (also called intensity) of HSV format image is adjusted by the histogram equalization process to enhance the image brightness. Thirdly, skin color detection and binarization process is employed on the resulting HSV image. Fourthly, the system dilates the binary image to make up the flaws on the image. Finally, the image is checked to determine if there is any possible skin color area which its horizontal width is larger than a threshold. If so, then the area is qualified to be a suspected object and the surveillance system controls a 2-dimentional rotation machinery to track the object. Meanwhile, the captured object image is synchronously displayed on a VGA screen and saved into a SD card. Designed under hardware and software co-design methodology, the resulting system has been experimentally proved that it satisfies the goal of real-time detecting and tracking of suspected objects.