An embedded system for vehicle surrounding monitoring

• Main Authors: Yuan-Yao Tu, 涂淵耀
• Other Authors: Chen, Yong-Sheng
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/72906890376298906319

碩士 === 國立交通大學 === 資訊科學與工程研究所 === 97 === What surrounds a vehicle effects vehicle maneuvering. Since there are a lot of blind spots around a vehicle, which lead to difficult maneuvering and endanger passengers and pedestrians. In this thesis, we develop a low cost but efficiency driving assistant system which provides the surrounding image of a vehicle in bird’s-eye view. By using a DSP chip and a fast-image-stitching algorithm as well as 4 fisheye cameras mounted around a vehicle, this system can instantly generate a vertical view of the vehicle from the top. This enables the driver to have a bird’s-eye view of their car and its surroundings without any blind spots. We also propose a novel idea to ensure the obstacle’s figure will show on the bird’s-eye view image. This vehicle surrounding monitoring system involves two major techniques, one is to find out the pixel mapping relationship between fisheye cameras and bird’s-eye view image, and the other is the real time image processing in embedded system. We propose some simple methods to find out the mapping relationship, including the fisheye image distortion model, fisheye image warping and rectifying. Then, we generate a lookup table for speeding up image processing in embedded system. On the other hand, we utilize the memory hierarchical structure and apply pipeline mechanism to enhance the throughput of embedded system. We also implement a dynamic boundary idea to let the system automatically switch image sources and reduce probability of missing obstacle’s figure in bird’s-eye view image. Finally, a bird’s-eye view image is generated. With this bird’s-eye view surrounding monitoring system, drivers can quickly understand the surrounding environment around the vehicle. Driving and parking become more easily and safely.