Font Size: a A A

Parking-Assistant System Using Omnidirectional Vision

Posted on:2011-08-20Degree:MasterType:Thesis
Country:ChinaCandidate:X DingFull Text:PDF
GTID:2178360302983124Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
A flexible calibration algorithm and a omnidirectional-vision parking assistant divice were proposed to generate a virtual bird -eye view image to assist the drivers in parking in the crowded cities.Four fish-eye cameras were fixed around the vehicle to capture images. At the beginning, fish-eye camera is calibrated to obtain the parameters and removed the fish-eye distortion, then the virtual bird-view point position was set in order to transform images into bird-view by using homography matrix. Stitching seam was labeled in overlapped area and also world coordinates frame were fixed; With stitching seam optimized and multi-bird-view images registered, the calibration parameters were output finally.After that, we described the design and implementation of the whole system. A TI TMS320DM643 was selected as the heart of that system and most of the control logical was implemented in a CPLD chip. The connection of chips was described in detail in the following. The parking assistant system included video recorders,video synchronous processor, DSP processor ,external memory,logic controller,output frame buffer, video coder and image display. The virtual bird-eye image was generate by the DSP processor which searched the look-up table of the virtual bird-eye image .This method requires only simple calculation and processes easily which has a good prospect of application.The last chapter drew a conclusion of the system and made some suggestion to the next version.
Keywords/Search Tags:omnidirectional-vision image, fish-eye calibration, homography matrix, stitching seam optimization
PDF Full Text Request
Related items