2016 16th International Conference on Control, Automation and Systems (ICCAS 2016)

Oct. 16-19, 2016 in HICO, Gyeongju, Korea

Histogram-model based Road Boundary Estimation by using Laser Scanner

Minho Cho, Jhonghyun An, Wonje Jang and Euntai Kim1*
1 Department of Electrical Electronic Engineering, Yonsei University, Seoul, Korea
([email protected], jhonghyen @yonsei.ac.kr, [email protected], [email protected])
* Corresponding author

Abstract: In this paper, we present road boundary estimation based on laser scanner by using histogram model. For raw
laser scanner data, we apply histogram model, which means that the histogram of the number of laser point is applied
from a center of ego-vehicle. After histogram is made, we estimate the boundary point and select point of road
boundary. We estimate selected points to clothoid model and update them by using Kalman filter. Experimental results
carried out with laser range data illustrate the robustness of our approach.

Keywords: laser scanner, road boundary, Kalman filter, clothoid model

1. INTRODUCTION
For intelligent transportation systems, advanced
driver assistance systems (ADAS) play very important
role in many interesting problems. They are automatic
cruise control, parking assistance, collision avoidance
systems, etc. In these domains, perception of the
environment and navigation of autonomous vehicles is
crucial to their efficiency. Among them, recognizing
drivable region is effective to economize the calculation
time and guarantees safety to driver. Up to now, there
are many researches to detect road boundaries and
surroundings, by using cameras, radar, and light
detection and raging (LIDAR) sensors. Stereo-vision
(a)
and camera are used to detect lane marks or road
boundaries [1]-[5] and LIDAR is also used to detect
road boundaries [6]-[9]. There is sensor fusion method
for road boundary detection [10]-[11]. For unusual
method, there are some researches by using grid
occupancy [12]-[13]. For detection road boundary, there
are some things to concern. First, it can measure and
classify surrounding environment precisely because
road boundary is predicted from estimated data, so if
system judged moving vehicles as road boundary, it (b)
cause critical problem. Moreover, it has to track or Figure 1 (a) raw data of laser scanner (b) road situation
update road boundary more stably. General road is
consecutive and parallel left and right boundary, so the
example, the raw data of laser scanner is shown as
more system can track the variation of road boundary,
figure 1-(a) and the road situation as figure 1-(b). For
the more precise estimation.
this raw laser scanner data, we make histogram to the
In this paper, we propose another method to estimate
center of ego-vehicle. The histogram of figure 1-(a) is
of road boundary by using only laser scanner. We apply
shown in figure 2. The y axis of histogram means the
histogram model, which means that the histogram of the
number of points of laser scanner for such x axis value
number of laser point is applied from a center of
and x value of that is same as the coordinate of
ego-vehicle. After histogram is made, we estimate the
ego-vehicle. In figure 2, there are the most number of
boundary point and select point of road boundary.
points in the 10m leftward from ego-vehicle and about
Finally we can estimate road boundary by using found
15m rightward from ego-vehicle so we can choose
points in previous and update them by using Kalman
boundary candidates from that area. Among those points
filter.
we choose boundary candidate by equal axis space
along y axis from designated x axis value-left and right,
2. BOUNDARY CANDIDATE SELECTION respectively- obtained from histogram. The result of
For raw laser scanner data, we apply histogram model, selected boundary candidate is shown as figure 3. The
which means that the histogram of the number of laser blue dots mean left boundary candidate, and black dots
point is applied from a center of ego-vehicle. For mean right boundary candidate, respectively.

1653
 In equation (2), xol and xor is left and right offset
of road boundary, respectively. These parameters are
updated by using Kalman filter and detailed procedure
is referred from [14].

4. EXPERIMENT
4.1 Experiment Setup
In this experiment, Scala 1403 (Valeo, 2D 4-layer) is
used mounted in front of vehicle. It has up to 145◦
horizontal field of view and the angular resolution is up
to 0.25◦. One unusual characteristic of Scala sensor
compared to another laser scanner is that scans three
layers at one period but does it twice, upper and lower.
The scanners are integrated at the front bumper of
testing vehicles as shown in figure 3.

Figure 2 Histogram of laser scanner point

Figure 4 Experiment setup

4.2 Experiment Results

The experiment is conducted various road
environment and results are shown in figure 5 and 6. In
each figure, estimated road boundary is expressed in
black line.
Figure 3 Selected boundary candidate. The blue dots We verify our recognition rate by comparing one by
mean left boundary candidate, and black dots mean one real picture and estimated boundary. We investigate
right boundary candidate, respectively. 10 minute video (about 2000 pictures and laser scan
data) in each environments. In city road, it shows about
85 percent of recognition rate and 90 percent of
recognition in highway. Maximum detection range of
3. ROAD BOUNDARY ESTIMATION road boundary is 50m in the front, 25m in left and right
In this research, we assume that the road is parallel. from the ego-vehicle.
Therefore, after selecting point for road boundary, we
exploit clothoid model to estimate the road boundary.
The equation of clothoid model is
2 3
y y
x ( y ) = c0 + xo + c1 (1)
22
In equation (1), x and y is coordinate of scan data,
xo is right and left offset of road bounday. For right
and left point of road boundary, the parameter is
updated by using Kalman filter. The state vector of road
boundary is Figure 5 Experiment result in city road
x = [c0 , c1 , xol , xor ] (2)

1654
 Road Boundary and Obstacle Detection Using a
Downward-Looking LIDAR Sensor”, IEEE Trans.
Intell. Transp. Syst., vol. 61, no. 3, pp. 971–985,
Mar. 2012.
[7] Y Kang, C Roh, SB Suh, B Song, “A lidar-based
decision-making method for road boundary
detection using multiple kalman filters”, IEEE
Trans. Idustiral Electronics, vol. 59, no. 11, pp.
4360–4368, Nov. 2012. .
[8] W Zhang, “LIDAR-based road and road-edge
detection”, Intelligent Vehicles Symposium (IV),
San Diego, June 2010.
Figure 6 Experiment result in highway [9] W. S. Wijesoma, K. R. S. Kodagoda, and A. P.
Balasuriya, “Road-Boundary Detection and
5. CONCLUSION Tracking Using Ladar Sensing”, IEEE Trans.
Robotics and automation, vol. 20, no. 3, pp.
This paper proposes road boundary estimation 456–464, Jun. 2004.
algorithm by using laser scanner. We conduct histogram [10] W.S. Wijesoma K.R.S. Kodagoda A. P. Balasuriya
model to select road boundary point. After selecting E.K.Teoh, “Road Edge and Lane Boundary
point, we estimate them to clothoid model and update Detection using Laser and Vision”, in Proc. IEEE
by using Kalman filter and it shows good performance Conf. Robot. Syst, Hawaii, USA, Nov 2001.
in various environments. [11] C. Rasmussen, “Combining Laser Range, Color,
In future work we plan to reinforce our road and Texture Cues for Autonomous Road
boundary modeling by using sensor fusion such as Following”, in Proc. IEEE Conf. Robot.
stereo vision and radar. Automation, May 2002.
[12] T. Weiss, B. Schiele and K. Dietmayer, “Robust
ACKNOWLEDGEMENT Driving Path Detection in Urban and Highway
This work was supported by the Technology Innovation Scenarios Using a Laser Scanner and Online
Program, (No. 10052731, Development of low level Occupancy Grids”in Proc. IEEE Conf. Intell.
video and radar fusion system for advanced pedestrian Vehicles Syst., Istanbul, Turkey, June 2007.
recognition) funded By the Ministry of Trade, industry [13] F. Homm, N. Kaempchen, J. Ota and D. Burschka,
& Energy(MOTIE), Korea. “Efficient Occupancy Grid Computation on the
GPU with Lidar and Radar for Road Boundary
Detection”, Intelligent Vehicles Symposium (IV),
6. REFERENCES
San Diego, June 2010.
[1] S.-J. Tsai and T.-Y. Sun, “The Robust and Fast [14] A. Kirchner and T. Heinrich, “Model based
Approach for Vision-based Shadowy Road detection of road boundaries with a laser scanner,”
Boundary Detection”, Proceedings of the 8th in Proc. of IEEE Intelligent Vehicles Symposium,
International IEEE Conference on Intelligent Stuttgart, Germany, 1998.
Transportation Systems, Vienna, Austria,
September 2005.
[2] M. Bertozzi and A. Broggi, “GOLD: a parallel
real-time stereo vision system for generic obstacle
and lane detection,” IEEE Trans. Image Process.,
vol. 7, no. 1, pp. 62–81, Jan. 1998.
[3] J. Crisman and C. Thorpe, “UNSCARF: a color
vision system for the detection of unstructured
roads,” in Proc. IEEE Conf. Robot. Automation,
1991, pp. 2496–2501.
[4] M. A. Sotelo, F. J. Rodriguez, and L. Magdalena,
“VIRTUOUS: Vision based road transportation for
unmanned operation on urban-like scenarios,”
IEEE Trans. Intell. Transp. Syst., vol. 5, no. 2, pp.
69–83, Jun. 2004.
[5] J. M. Alvarez, T. Gevers, A. M. Lopez,
“VISION–BASED ROAD DETECTION USING
ROAD MODELS”, Image Processing (ICIP),
2009 16th IEEE International Conference, Nov.
2009.
[6] J. Han, D. Kim, M. Lee, M. Sunwoo, “Enhanced

1655