Method

Binary Map from color-based road detection [st] [BM]


Submitted on 20 Jun. 2014 16:38 by
Bihao WANG (Université de Technologie de Compiègne)

Running time:2 s
Environment:2 cores @ 2.5 Ghz (Matlab)

Method Description:
An image conversion into a log-chromaticity space
provides an efficient and stable identification
(i.e. illuminant invariant) of road surface
intrinsic features. Then, sample pixels are
randomly selected from an assumed “road” area.
Next, a confidence interval is defined using the
samples to classify pixels of the intrinsic image
into the road and nonroad surfaces. Finally,
stereo-vision based extension grants access to the
3D road profile estimation and enhances the
detection precision.
Parameters:
confidence level 1-\alpha=0.75
Latex Bibtex:
@inproceedings{Wang2014IVWORK,
author={WANG, B. and Fremont, V. and Rodriguez
Florez, S. A.},
title={Color-based Road Detection and its
Evaluation on the KITTI Road Benchmark},
booktitle={Workshop on Benchmarking Road Terrain
and Lane Detection Algorithms for In-Vehicle
Application, IEEE Intelligent Vehicles Symposium},
year={2014},
pages={31-36},
}

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 78.90 % 66.06 % 69.53 % 91.19 % 18.21 % 8.81 %
UMM_ROAD 89.41 % 80.61 % 83.43 % 96.30 % 21.02 % 3.70 %
UU_ROAD 78.43 % 62.46 % 70.87 % 87.80 % 11.76 % 12.20 %
URBAN_ROAD 83.47 % 72.23 % 75.90 % 92.72 % 16.22 % 7.28 %
This table as LaTeX

Behavior Evaluation


Benchmark PRE-20 F1-20 HR-20 PRE-30 F1-30 HR-30 PRE-40 F1-40 HR-40
This table as LaTeX

Road/Lane Detection

The following plots show precision/recall curves for the bird's eye view evaluation.


Distance-dependent Behavior Evaluation

The following plots show the F1 score/Precision/Hitrate with respect to the longitudinal distance which has been used for evaluation.


Visualization of Results

The following images illustrate the performance of the method qualitatively on a couple of test images. We first show results in the perspective image, followed by evaluation in bird's eye view. Here, red denotes false negatives, blue areas correspond to false positives and green represents true positives.



This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png


eXTReMe Tracker