ZeePedia

Advanced Topics

<< Planning
Conclusion >>
img
Artificial Intelligence (CS607)
9 Advanced Topics
9.1 Computer vision
It is a subfield of Artificial Intelligence. The purpose of computer vision is to study
algorithms, techniques and applications that help us make machines that can
"understand" images and videos. In other words, it deals with procedures that
extract useful information from static pictures and sequence of images. Enabling
a machine to see, percieve and understand exactly as humans see, percieve and
understand is the aim of Computer Vision.
Computer vision finds its applications in medicine, military, security and
surveillance, quality inspection, robotics, automotive industry and many other
areas. Few areas of vision in which research is benig actively conducted
thoughout the world are as follows:
The detection, segmentation, localisation, and recognition of certain
objects in images (e.g., human faces)
Tracking an object through an image sequence
Object Extraction from a video sequence
Automated Navigation of a robot or a vehicle
Estimation of the three-dimensional pose of humans and their limbs
Medical Imaging, automated analysis of different body scans (CT Scan,
Bone Scan, X-Rays)
Searching for digital images by their content (content-based image
retrieval)
Registration of different views of the same scene or object
Computer vision encompases topics from pattern recognition, machine learning,
geometry, image processing, artificial intelligence, linear algebra and other
subjects.
Apart from its applications, computer vision is itself interesting to study. Many
detailed turorials regarding the field are freely avalible on the internet. Readers of
this text are encouraged to read though these tutorials get indepth knowledge
about the limits and contents of the field.
Exercise Question
Search though the internet and read about intersting happeneing and reseach
going on around the globe in the are of Computer Vision.
203
img
Artificial Intelligence (CS607)
http://www.cs.ucf.edu/~vision/
The above link might be useful to explore knowledge about computer vision.
9.2 Robotics
Robotics is the highly advanced and totally hyped field of today. Literally
speaking, robotics is the study of robots. Robots are nothing but a complex
combination of hardware and intelligence, or mechanics and brains. Thus
robotics is truly a multi-disciplinary area, having active contributions from,
physics, mechanics, biology, mathematics, computer science, statistics, control
thory, philosophy, etc.
The features that constitute a robot are:
Mobility
·
Perception
·
Planning
·
Searching
·
Reasoning
·
Dealing with uncertainty
·
Vision
·
Learning
·
Autonomy
·
Physical Intelligence
·
What we can see from the list is that robotics is the most profound manifestation
of AI in practice. The most crucial or defining ones from the list above are
mobility, autonomy and dealing with uncertainety
The area of robotics have been followed with enthusiasm by masses from fiction,
science and industry. Now robots have entered the common household, as robot
pets (Sony Aibo entertainment robot), oldage assistant and people carriers
(Segway human transporter).
Exercise Question
Search though the internet and read about intersting happeneing and reseach
going on around the globe in the are of robotics.
http://www.cs.dartmouth.edu/~brd/Teaching/AI/Lectures/Summaries/robotics.html
The above link might be useful to explore knowledge about robotics.
9.2.1 Softcomputing
Softcomputing is a relatively new term coined to encapsulate the emergence of
new hybrid area of work in AI. Different technologies including fuzzy systems,
204
img
Artificial Intelligence (CS607)
genetic algorithms, neural networks and a few statistical methods have been
combined together in different orientations to successfully solve today's complex
real-world problems.
The most common combinations are of the pairs
·  genetic algorithms ­ fuzzy systems (genetic fuzzy)
·  Neural Networks ­ fuzzy systems (neuro-fuzzy systems)
·  Genetic algorithms ­ Neural Networks (neuro-genetic systems)
Softcomputing is naturally applied in machine learning applications. For example
one usage of genetic-fuzzy system is of `searching' for an acceptable fuzzy
system that conforms to the training data. In which, fuzzy sets and rules
combined, are encoded as individuals, and GA iterations refine the individuals
i.e. fuzzy system, on the basis of their fitness evaluations. The fitness function is
usually MSE of the invidual fuzzy system on the training data. Very similar
applications have been developed in the other popular neuro-fuzzy systems, in
which neural networks are used to find the best fuzzy system for the given data
through means of classical ANN learning algorithms.
Genetic algorithms have been employed in finding the optimal initial weights of
neural networks.
Exercise Question
Search though the internet and read about intersting happeneing and reseach
going on around the globe in the are of softcomputing.
http://www.soft-computing.de/
The above link might be useful to explore knowledge about softcomputing.
9.3 Clustering
Clustering is a form of unsupervised learning, in which the training data is
available but without the classification information or class labels. The task of
clustering is to identify and group similar individual data elements based on some
measure of similarity. So basically using clustering algorithms, classification
information can be `produced' from a training data which has no classification
data at the first place. Naturally, there is no supervision of classification in
clustering algorithms for their learning/clustering, and hence they fall under the
category of unsupervised learning.
The famous clustering algorithms are Self-organizing maps (SOM), k-means,
linear vector quantization, Density based data analysis, etc.
Exercise Question
205