Research On Vehicle Carried Vision Perseption System Based On FPGA

People have been looking for a more secure, faster and more convenient transportation for decades, to solve the problem of low efficiency, high accident rate and driving difficulties. With the intensive development of modern electronic technology and automatic control algorithms, intelligent vehicle gradually turns concept into reality.The intelligent vehicle obtains the information of the traffic environment and the status of the vehicle itself by many kinds of sensors installed on the vehicle to perceive the environment all around. The intelligent vehicle make judgments and decisions, plan travel path and operate the vehicle by analyzing the perception results. Most traffic signals and road environment information is transmitted to the driver through a visual way, similarly machine vision shall be the key component of environment perception device on intelligent vehicle.In recent years, foreign research institutions and automobile manufacturers put great efforts on intelligent vehicle research and development and have been making intelligent driving assist system more and more practical. There is a great gap between domestic and foreign research on the related field, so it is significant to research and develop intelligent vehicle vision system with independent intellectual property rights.In view of the current status of domestic and foreign research in machine vision on intelligent vehicle, a distributed intelligent vehicle carried multi-camera vision system based on FPGA processors has been designed and tested in this thesis to fusion multiple cameras captured image data and overcome the installation difficulties. Four CMOS image sensor capture and processing boards installed on the sides of vehicle, with a main control board which is used for video data fusion and storage formed the vehicle carried intelligent vision system.The vision system used FPGA processor to capture video in a parallel way. High speed digital serial interfaces based on LVDS were used to transmit data and commands. A DDR2 SDRAM was used to store large volume video data rapidly. This vehicle carried embedded hardware platform made the video fusion function capable, reduced the amount of cable, and offered flexible camera interfaces.It is significant to identify the lane line when intelligent vehicle perceiving the environment. An orange lane detection and extraction method based on image processing algorithms has been designed and tested. The image processing experiment showed good extraction result under different light and weather conditions by transforming the image format from RGB into HSV and applying OTSU algorithm to reduce the impact of light conditions.The method has been implemented on FPGA platform in an optimized way using parallel computing and pipelining methodology. The calculation process has been fully optimized to meet the requirement of both speed and area utilization. Compared to other serial execution embedded processors, this method has a good algorithm acceleration effect. The presented solution is suitable for vehicle carried intelligent vision applications.The designed system has high intergration level and computing performance, can be used in small installation space and high processing speed required ADAS applications, which is an economical and practical solution. To our knowledge, the presented distributed embedded vision system based on FPGA which integrated with accelerated image processing algorithm has not been reported in China.