Você está na página 1de 3

Chapter 4 Discussion and Future Work

Chapter 4 Discussion and Future Work

4.1 Discussion

After working endlessly and doing unlimited studies on this final year project, it can
conclude that the project objective, to develop an algorithm that will identify road signs
from real time camera, video file and static digital images in a reasonable time frame, has
been met with success.
As this project needs a large number of image databases and road sign examples to make
system responding correctly, it is a time-consuming task; the project also involves mainly
on the software development and I did not have any program development experience
before, its a big challenge to complete it. But through the process of the development, I
have touched MATLAB program which include Image Processing, Neuron Network and
GUI toolboxes, and learned a lot for the software development management skills.

4.2 Future Work

The most important for future work to be done for present project is to collect large number
of road sign images, expanding database, train the neuron network again and modify the
classification program. And make it able to identify another two type road signs, No Entry
and Speed Limit signs. At the same time, the system performance could be tested.
However there exist possibilities to further enhance the research work which can lead to
further improvement of functionality and robustness of the algorithms proposed. This
section suggests a number of possible enhancements by introducing new software
toolboxes, image capturing devices, color correction algorithms, object tracking
algorithms, advance feature selectors and classifiers.

30
Chapter 4 Discussion and Future Work

1. The implementations of proposed algorithms have been carried out by using


different programming platforms such as MATLAB, C/C++, C# and Java. The
implementations of color space transformation of an image is carried out in C#
which has significantly reduced the processing time as compared to algorithms
previously implemented in MATLAB. One of the future tasks is to transform all
algorithms into one software development environment such as C# that has the
capability to increase the speed of execution and processing of the TSDR system as
desired by real time applications. It is believed that the conversion of the MATLAB
implementation to C# has the ability to increase the speed of algorithms by up to
eight times.

2. The performance of the proposed color segmentation techniques still need to be


compared with other classical segmentation techniques such as clustering, edge
detection, and region based methods, vector based methods and fuzzy techniques
as well as physics based methods.

3. The position and installation of image acquisition device in the car is also very
important to improve the accuracy of detection and recognition of the road signs.
Recently car manufactures have included the speed limit sign detection systems in
the cars and the camera installation point is selected as the back side of the
windscreen mirror, to avoid the possible shake during image acquisition.

4. In future work the study of dual camera based image acquisition systems can be
investigated with careful design of their position and installation within the car, to
obtain meaningful information (including depth) as well as helping to avoid
possible shake and vibration during image acquisition.

5. The present state of the art and the fundamental theories of computer vision provide
a large number of feature detectors which can be combined with large number of
advance classifiers to obtain the best recognition accuracy.
31
Chapter 4 Discussion and Future Work

It is highly recommended that research shall be carried out to determine the use of
appropriate feature descriptors and classifiers which are not listed and investigated within
this thesis. It is an open research problem to find the best possible feature descriptor and
classifier pair that gives optimum accuracy.

32

Você também pode gostar