The KITTI Vision Benchmark Suite

Method

Appearance-Aligned Relaxed Block-Matching Stereo and RANSAC plane estimation [st] [SPlane]

Submitted on 23 May. 2014 14:04 by
Nils Einecke (Honda Research Institute Europe GmbH)

Running time:		2 s
Environment:		1 core @ 3.0 Ghz (C/C++)

Method Description:

1.) Disparity calculation using AARBM stereo method
2.) RANSAC plane estimation
3.) Subtract plane from disparity map to generate a road confidence map

You can request a copy of the AARBM stereo paper here.

Parameters:

disp.-range = [0..255]
BM-size = 11x11
RBM penalty = 0.9

Latex Bibtex:

@InProceedings{Einecke2014IV,
author = {N. Einecke and J. Eggert},
title = {Block-Matching Stereo with Relaxed Fronto-Parallel Assumption},
booktitle = {IV},
pages = {700--705},
year = 2014
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	78.19 %	76.41 %	72.03 %	85.50 %	15.13 %	14.50 %
UMM_ROAD	82.28 %	82.83 %	76.85 %	88.53 %	29.32 %	11.47 %
UU_ROAD	73.30 %	69.11 %	65.39 %	83.38 %	14.38 %	16.62 %
URBAN_ROAD	78.69 %	77.16 %	71.96 %	86.80 %	18.63 %	13.20 %

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.

This figure as: png eps pdf

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