Research On Bird’s-eye Panoramic View System Based On Image Quality Assessment

The bird’s-eye panoramic view system based on image processing technology can effectively eliminate the driver’s blind spot in driving and provide an effective solution to driving safety problems.At present,the bird’s-eye panoramic view system is widely used in luxury cars,but less used in vehicles.In this thesis,the objective method of image mosaic quality is introduced to design panoramic driving assistance system for trucks.The design and parameters of bird’s-eye panoramic view system are selected and optimized.The system is built on the embedded platform and the engineering application is studied.In this thesis,the existing image quality assessment methods are studied.The objective image quality assessment algorithm can be divided into full reference image quality assessment algorithm,semi-reference image quality assessment algorithm and non-reference image quality assessment algorithm according to existence or nonexistence of reference image.Aiming at the typical full reference image quality assessment algorithm and non-reference image quality assessment algorithm,this thesis carried out in-depth research and analysis,including MSE,PSNR,SSIM and other algorithms.In this thesis,a robust SIFT feature point extraction algorithm is used,and the RANSAC algorithm is applied to remove the point pairs.Based on the Gabor filter,the difference information sensing model for image mosaic quality is improved,the edge extraction of the mosaic image is made,and the mosaic image of the edge image is also made,then the difference image of the edge is obtained by combining with the JND.Finally,we use the evaluation index to calculate the evaluation results.The bird’s-eye panoramic view system includes many key parts,such as the fish-eye camera distortion correction algorithm,the camera calibration method,image fusion and brightness equalization method,the bird’s-eye panorama generation model.Considering the large size of truck,this thesis presents an automatic and rapid calibration method for external reference of cameras under large field of view based on X-shaped target,and improves a method of brightness equalization based on partial pixels in overlapping area.Based on the theoretical research,this thesis studies on the engineering application of bird’s-eye panoramic view system on truck.Based on the high performance ARM processor,an embedded hardware environment and a software development environment are constructed,and the embedded software development and algorithm migration of the bird’s-eye panoramic system are implemented,and the hardware and software resources are fully utilized to accelerate the system.In order to obtain the optimized system parameters,the evaluation model based on the edge image is used to evaluate the system scheme and parameter selection.Finally,the complete embedded bird’s-eye panoramic view system is realized and applied to the mine truck.