Open AccessVision‐based position computation from in‐vehicle video log images for road sign inventory
Author(s) -
Hu Zhaozheng,
Li Na
Publication year - 2016
Publication title -
iet intelligent transport systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.579
H-Index - 45
eISSN - 1751-9578
pISSN - 1751-956X
DOI - 10.1049/iet-its.2015.0078
Subject(s) - computer vision , artificial intelligence , computer science , computation , position (finance) , sign (mathematics) , camera resectioning , calibration , enhanced data rates for gsm evolution , field of view , mathematics , algorithm , economics , mathematical analysis , statistics , finance
Computing the positions of road signs from in‐vehicle video log images is one of the most fundamental tasks for intelligent sign inventory. This study analysed vision‐based positioning methods from video log images and proposed an analytic model of road sign positioning (AM‐RSP) based on error propagation and first‐order approximation. Both image noises and camera calibration errors were modelled as two error sources. The AM‐RSP model can evaluate and quantify the positioning uncertainties for both single view and stereo vision‐based methods. Moreover, it can quantitatively establish the relationship between the positioning uncertainties and the various parameters of the vehicle and the camera,such as camera focal length, view angle, baseline width, height, image acquisition interval, vehicle distance to the road edge etc. Hence, the proposed AM‐RSP model can be applied to configure the parameters of the camera and the vehicle to meet some specific computation requirement such as RSP accuracy. The proposed AM‐RSP model has been validated with real video log image data collected in the field. These results show that the AM‐RSP model is accurate and reliable to compute sign positioning uncertainties.