Real Time Vehicular Camera Vision Acquisition System Using Field Programmable Gate Array (FPGA)

With the advent of active safety technologies in the automotive industry, a need to record and replay the actual on-road vehicular scenario has risen, especially in systems involving camera-based vision. The primary objective of the paper is to propose a design of a system for real-time video acquisition. Hence, a design for a Camera Hardware simulator has been proposed in this paper. The system involves a camera that captures visual information through its image sensor. The system is designed such that it can do direct display; that is, it can generate vertical and horizontal synchronization signals, as per the specification of the camera and it can buffer the pixel clock coming from the camera and send it to another system that uses the video information being received such as an in-vehicle display to display it. It also includes the ability to record the incoming data stream in a computer for offline processing. As the aforementioned functionality is to be achieved for high incoming data rates and also handy interfacing to any recording device is required, we are using a Universal Serial Bus (USB) 2.0 (high speed). Many subsystems are to be designed on the same chip, so we propose the use of a Field Programmable Gate Array (FPGA) based system for fast data processing and miniaturization of the system. The system under consideration is comprised of a Complementary Metal-Oxide Semiconductor (CMOS) camera at the input and a high-speed USB interface at the output. The FPGA is programmed as a Video Graphics Array (VGA) Controller, buffer and a USB Controller. FPGA consumes less power when compared to any other embedded-based system, making it more usable inside an automobile. This system can be used for offline video processing and simulation of the exact on-road scenario in lab for vision based active safety system's testing purpose or for active safety algorithm improvement to achieve desired results (by tweaking the algorithm for desired results, based on the observations made from the recorded data without having the need to go on road again and again).