A real-time multipoint tracking system based on FPGA for multi-touch and motion tracking

Blob detection which focuses on detecting points or regions of a different intensity than the surrounding image is increasingly used in consumer products such as human-computer interfaces and motion tracking. Because blob detection is computationally intensive but requires relatively simple arithmetic operations, it is an ideal candidate for parallelization in hardware. The main goal of this paper is to develop a hardware implementation for blob detection structure that is able to detect multipoint in a video image on a Xilinx FPGA platform. This system consists of three functional blocks. The first block use a dual port memory to get the histogram of video data and then to obtain the threshold value for the image frame. The second block applies the threshold value to the video stream data, gets the line connected component, and these components are then transferred into the third block by Fast Simplex Link (FSL). The third block is a Microblaze processor which does the label connection and gets the center of points. This approach is implemented on a Xilinx Spartan3 chip with 640 by 480 resolutions up to 30Hz. It can be used in various low cost consumer applications.