Graphics Reference
In-Depth Information
Table 10.3
Design configurations by varying
n
B
and
w
B
n
B
ₓ
n
B
ₓ
n
B
ₑ
n
B
ₑ
w
B
ₓ
w
B
ₑ
w
B
ₓ
w
B
ₑ
Lane types
Solid
√
√
√
ₔ
√
Dashed
√
Circular reflectors
bot dots
Environmental
Sunny day
Depends on lane types above
√
a
Night scene
√
a
Foggy conditions
√
a
Rainy conditions
Embedded constraints
Parallel processing
√
√
√
√
Low area constraint
√
√
Pipelining
√
√
Low memory resources
Timing Depends on hardware configuration
a
This also depends on placement of scan bands
First, we consider the types of lane markings and what combination of scan band
size
w
B
and number
n
B
could give acceptable detection rates. For example, solid
lanes require minimal number of scan bands and can also work with smaller band
sizes. However, circular reflectors need higher number and wider band sizes also.
Similarly, certain combinations of
n
B
and
w
B
are suited for specific environmental
conditions. For example, in foggy and rainy conditions, it is desirable to extract lanes
from the road surface closest to ego-vehicle. Therefore, lesser number of bands but
wider bands closer to the vehicle are sufficient for robust extraction.
In the next part of Table
10.3
, we consider the different configurations for
n
B
and
w
B
that comply with certain embedded constraints/requirements. A combination of
the selections between the different categories can be used to give a user-constrained
embedded realization of an accurate lane feature extraction system.
10.5 Conclusions
In this paper, we proposed a lane extraction method that is shown to provide a
way to explore the different configurations for embedded realization, based on the
user requirements and the context in which the lane analysis is to be done. It is
shown that the two design parameters, i.e., number of scan bands and width of
