Intro to Remote Sensing

advertisement
Artificial Intelligence (AI)
Addition to the lecture 11
What is AI?


It is the science and engineering of making intelligent
machines, especially intelligent computer programs. It is
related to the similar task of using computers to understand
human intelligence, but AI does not have to confine itself to
methods that are biologically observable
Applications of AI






game playing
speech recognition
understanding natural language
computer vision
expert systems
heuristic classification
http://www-formal.stanford.edu/jmc/whatisai/node3.html
Knowledge-based expert system




Artificial neural network (ANN)
Decision tree
Support vector machines (SVMs)
…
Knowledge representation process
The knowledge representation process normally involves encoding
information from verbal descriptions, rules of thumb, images, books,
maps, charts, tables, graphs, equations, etc. Hopefully, the knowledge
base contains sufficient high-quality rules to solve the problem under
investigation. Rules are normally expressed in the form of one or
more “IF condition THEN action” statements. The condition portion
of a rule statement is usually a fact, e.g., the pixel under investigation
must reflect > 45% of the incident near-infrared energy. When certain
rules are applied, various operations may take place such as adding a
newly derived derivative fact to the database or firing another rule.
Rules can be implicit (slope is high) or explicit (e.g., slope > 70%). It
is possible to chain together rules, e.g., IF c THEN d; IF d THEN e;
therefore IF c THEN e. It is also possible to attach confidences (e.g.,
80% confident) to facts and rules.
For example, a typical rule used by the MYCIN expert system is
IF the stain of the organism is gram-negative
AND the morphology of the organism is rod
AND the aerobicity of the organism is anaerobic
THEN there is strong suggestive evidence (0.8) that the
class of the organism is Enterobacter iaceae.
Following the same format, a typical remote sensing rule might be:
IF blue reflectance is (Condition) < 15%
AND green reflectance is (Condition) < 25%
AND red reflectance is (Condition) < 15%
AND near-infrared reflectance is (Condition) > 45%
THEN there is strong suggestive evidence (0.8) that the
pixel is vegetated.
1. ANN



The motivation for the development of neural network technology stemmed from the
desire to develop an artificial system that could perform "intelligent" tasks similar to
those performed by the human brain (thousands of different inputs-neurons, output to
many other neurons), with

Simple processing elements

A high degree of interconnection

Simple scalar messages

Adaptive interaction between elements
ANN usually has one input layer, one output layer, and no or some hidden layers
between. Neurons in one layer are connected to all neurons in the next layer for passing
information
Neural networks process information in a similar way the human brain does. The network
is composed of a large number of highly interconnected processing elements (neurones)
working in parallel to solve a specific problem. Neural networks learn by example. They
cannot be programmed to perform a specific task. The examples must be selected
carefully otherwise useful time is wasted or even worse the network might be functioning
incorrectly. The disadvantage is that because the network finds out how to solve the
problem by itself, its operation can be unpredictable.
How do ANN work?

Train the Network





Input training sites to the network
Network computes an output
Network output compared to desired output
Network weights are modified to reduce error
Use the network


Input new data to the network
Network computes outputs based on its training
An example of a complicated ANN
2. Decision tree

"A decision tree takes as
input an object or situation
described by a set of
properties, and outputs a
yes/no decision. Decision
trees therefore represent
Boolean functions.
Functions with a larger
range of outputs can also
be represented...."
Cont’


A decision tree is a type of multistage classifier that can
be applied to a single image or a stack of images. It is
made up of a series of binary decisions that are used to
determine the correct category for each pixel. The
decisions can be based on any available characteristic of
the dataset. For example, you may have an elevation
image and two different multispectral images collected
at different times, and any of those images can
contribute to decisions within the same tree. No single
decision in the tree performs the complete segmentation
of the image into classes. Instead, each decision divides
the data into one of two possible classes or groups of
classes.
Image segmentation (eCognition) +
decision tree (such as see5 at
http://www.rulequest.com/see5-info.html)
Hierarchical Decision Tree Classifier
ETM Panchromatic
Expert’s Model
Predicted White Fir
Hierarchical Decision Tree
Classifier Based on
Inductive Machine
Learning Production Rules
ETM Panchromatic
C5.0 Model
Predicted White Fir
Machine Learning-derived Classification Map
Thomas, et al. 2003, PERS
Cont’


ENVI’s decision tree tool is designed to
implement decision rules, such as the
rules derived by any number of excellent
statistical software packages that provide
powerful and flexible decision tree
generators. Two examples that are used
commonly in the remote sensing
community include CART by Salford
Systems and S-PLUS by Insightful. The
logic contained in the decision rules
derived by these software packages can
be used to build a decision tree classifier
with ENVI’s interactive decision tree
tool.
Even if you have not used one of these
packages to derive any decision rules,
you may find ENVI’s new decision tree
tool to be a useful way to explore your
data, or to find areas in your data that
fulfill certain criteria.
3. Support vector machines (SVMs)



Is a new generation learning system based on recent
advances in statistical learning theory
SVMs deliver state-of-the-art performance in realworld applications such as text categorisation, handwritten character recognition, image classification,
biosequences analysis, etc.
SVMs’s first introduction in the early 1990s lead to a
recent explosion of applications and deepening
theoretical analysis, that has now established SVMs
along with neural networks as one of the standard
tools for machine learning and data mining
Want to learn more?






http://svmlight.joachims.org/
http://svm.dcs.rhbnc.ac.uk/
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
http://theoval.sys.uea.ac.uk/~gcc/svm/toolbox/
http://www.cs.wisc.edu/dmi/lsvm/
http://vision.ai.uiuc.edu/mhyang/svm.html
Download