The KITTI Vision Benchmark Suite

Method

CRF based Road Detection with Multi-Sensor Fusion[la] [FusedCRF]

Submitted on 30 Jan. 2015 02:27 by
Liang Xiao (National University of Defense Technology)

Running time:		2 s
Environment:		1 core @ 2.5 Ghz (C/C++)

Method Description:

First align the laser points and the pixels. Then learn
the boosted decision tree based pixel classifier and
point cloud classifier. Using the classifiers scores as
the potentials of the CRF and jointly optimized with
graph cut.

Parameters:

Boosting: K = 50 rounds with tree depth=4.
CRF: lambda = 10, gamma = 45;

Latex Bibtex:

@inproceedings{Xiao2015IV,
author = "Liang Xiao and Bin Dai and Daxue Liu and
Tingbo Hu and Tao Wu",
title = "CRF based Road Detection with Multi-Sensor Fusion",
booktitle = "Intelligent Vehicles Symposium (IV)",
year = "2015",
}

Evaluation in Bird's Eye View

Benchmark	MaxF	AP	PRE	REC	FPR	FNR
UM_ROAD	89.55 %	80.00 %	84.87 %	94.78 %	7.70 %	5.22 %
UMM_ROAD	89.51 %	83.53 %	86.64 %	92.58 %	15.69 %	7.42 %
UU_ROAD	84.49 %	72.35 %	77.13 %	93.40 %	9.02 %	6.60 %
URBAN_ROAD	88.25 %	79.24 %	83.62 %	93.44 %	10.08 %	6.56 %

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