Research And Design Of Vehicle-mounted Panorama System

Nowadays the car ownership in China is growing rapidly,and yet there is still much room for improvement.Cars can provide much convenience to people,but also cause traffic security issues.Accidents happen frequently due to the blind spot when driving,causing unnecessary losses.Vehicle-mounted panorama system is a brand-new driving assist system in recently years,which can provide real-time panoramic images around the car to the driver.It solves the trouble caused by the blind spot and makes driving much safer.The vehicle-mounted panorama system algorithm has been researched and tested in sequence in this thesis.Firstly,original images taken by fisheye cameras are calibrated to remove the distortion effect,and turned into top view images by using inverse perspective transformation.All the images are stitched into the panorama image by the stitching method,and an image dodging algorithm is designed to cope with the brightness and color difference between images.The hardware processing flow is optimized to improve instantaneity and stability of the system.The following works are included to fulfill this system:(1)To obtain the intrinsic parameters and distortion coefficient of the fisheye lens,the calibration test is conducted by Zhang calibration method.We use the parameters to correct the distortion and inverse the perspective by bilinear interpolation to get top view images.(2)Image registration algorithm based on local feature and RANSAC are studied and tested,and we analyzed its failure in this system.A new image registration method is proposed based on the invariance of the relative position of cameras.(3)An image dodging algorithm is designed to solve the brightness and color difference as well as the obvious seams caused by illumination unbalance,and the image quality is improved.(4)The experiment is designed and conducted.Images are taken and processed into panoramic image based on the research.The hardware architecture is designed and its processing flow is optimized by loading the look up table to improve the processing efficiency and user experience.The system researched and designed in this thesis can realize seamless panoramic image,solve the brightness and color difference caused by illumination unbalance and improve the processing efficiency by processing flow optimization.With a fine image quality and stability to satisfy the demands of usage,this system has a widely application prospect.