Font Size: a A A

Design And Implementation Of Dm8148-Based Panoramic Mosaic Algorithms And Video Recording Software

Posted on:2015-11-04Degree:MasterType:Thesis
Country:ChinaCandidate:Z L WangFull Text:PDF
GTID:2298330467954942Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
As there are a lot of objective reasons, such as too many vehicles and people, the lack of parking places, vision limitation in vehicles and so on, which make the traditional driver assistance system of vehicles could not satisfy the citizens’ expectation for safety driving. Recently, the performance of vehicle processing devices continues improving and the technology of video encoding develops very fast which make vehicles own panoramic driver assistance system possibly. Panoramic driver assistance system can provide the view of whole scene around the vehicle. It will make drivers to avoid vehicle collision and park in a heavy environment easily.For the presence of the blind region in vehicles, inconvenience of the multi-channel video surveillance and lack of video record, this thesis designs and implements a panoramic bird-eye image mosaic algorithm and a video recording software based on the OpenMax framework. Most of work and achievements are presented as follows.(1) After analyzing the common calibration methods for cameras and the differences between the common cameras and fish-eye cameras, this thesis designs a calibration method for fish-eye cameras by using Least Square method. Designing and implementing the fish-eye image correction algorithm for correcting the distortion of fish-eye images based on the sphere projection model. The experimental results show that the fish-eye image correction algorithm can effectively correct the distortion of fish-eye images.(2) After analyzing the relation between the image’s coordinates and the real world’s coordinates for perspective projection, utilizing the homogeneous linear equations to solve image rotation transform matrix and combining it with image interpolation algorithm to implement virtual overlook-view transformation algorithm. The experimental results show the virtual overlook-view transformation algorithm effectively turns side-look images into over-look images. (3) Considering that4images processed by fish-eye image correction algorithm and virtual overlook-view transformation algorithm have deformation and distortion which makes common image mosaic algorithm can’t be applied to the panoramic bird-view image mosaic algorithm, this thesis designs and implements an image mosaic algorithm based on the projection relation between panoramic image coordinates and ground coordinates. Experimental results show that the panoramic bird-view image processed by the panoramic bird-view image mosaic algorithm can help drivers watch the obstacles in the blind region.(4) The panoramic bird-view image mosaic algorithm is implemented and optimized for the high parallel computing capacity of float-point C674x DSP on DM8148platform. By utilizing the OpenMax framework from TI supports, a video recording software is implemented for encoding and saving the panoramic video. Experimental results show the DM8148-based video recording software can encode the panoramic bird-view video in real time.
Keywords/Search Tags:panoramic bird-view surveillance, fisheye image correction, image mosaic, H.264codec, DM8148
PDF Full Text Request
Related items