This document is a seminar report submitted by Ms. Nidhi V. Shirbhayye to the Department of Electronics and Telecommunication Engineering at Shri Sant Gajanan Maharaj College of Engineering in Shegaon, India. The report proposes a new method for traffic density estimation based on topic modeling, which can automatically discover motion patterns in traffic scenes using unsupervised learning. It reviews previous work on traffic monitoring using magnetic loop detectors, video surveillance systems, and computer vision. The report will describe the background theory and methodology of the proposed topic modeling approach for accurate, real-time traffic density estimation.

1. i
SSGMCE, Shegaon
A
Seminar Report
On
A New Method for Traffic Density Estimation
Based on Topic Model
Submitted in partial fulfillment of
the requirements for the degree of
Bachelor of Engineering
in
Electronics & Telecommunication Engineering
of
Sant Gadge Baba Amravati University, Amravati
Submitted by
Ms. Nidhi V. Shirbhayye
(Class & Roll No.:4U1-14)
Under the esteemed guidance of
Prof. A. N. Dolas
Asst. Prof., E & TC Dept.
Department of Electronics & Telecommunication Engineering
Shri Sant Gajanan Maharaj College of Engineering, Shegaon,
Dist- Buldhana – 444 203 (Maharashtra)
(2016-17)

2. ii
SSGMCE, Shegaon
Certificate
The seminar report entitled “A New Method for Traffic Density Estimation
based on Topic Model- Current status and Future Perspectives” is hereby
approved as a creditable study carried out and presented by Ms. Nidhi V. Shirbhayye
in a manner satisfactory to warrant its acceptance as a pre-requisite in a partial
fulfillment of the requirements for degree of Bachelor of Engineering in Electronics
& Telecommunication Engineering of Sant Gadge Baba Amravati University,
Amravati.
Prof. A. N. Dolas Dr. G. S. Gawande
Guide Prof. & Head, E & TC Dept.
Internal Examiner
Department of Electronics & Telecommunication
Shri Sant Gajanan Maharaj College of Engineering,
Shegaon – 444203, Maharashtra, India

3. iii
SSGMCE, Shegaon
Abstract
Traffic density estimation plays an integral role in intelligent transportation
systems (ITS) for controlling signals. It provides important information in ITS for
road planning, intelligent road routing, effective traffic management, road traffic
control, network traffic scheduling, routing and dissemination. The system presents a
new framework for traffic density estimation based on topic model, which is an
unsupervised model. It uses a set of visual features without any individual vehicle
detection and tracking need, and discovers the motion patterns automatically in traffic
scenes by using topic model. Then, likelihood value allocated to each video clip
enables us to estimate its traffic density.It shows high classification performance and
robustness to typical environmental and illumination conditions and estimates the
density of traffic videos even in bad illumination condition. Since the execution time
for this approach is relatively low, it can be used in real-time application.

4. iv
SSGMCE, Shegaon
Acknowledgement
The real spirit of achieving a goal is through the way of excellence and lustrous
discipline. I would have never succeeded in completing my task without the
cooperation, encouragement and help provided to me by various personalities. There
are a number of people who deserve recognition for their unwavering support and
guidance throughout this report.
I am highly indebted to my guide Prof. A. N. Dolas for his guidance and
constant supervision as well as for providing necessary information regarding the
report & also for their support in completing the report. I would like to take this
opportunity to express my heartfelt thanks, for his esteemed guidance and
encouragement. His suggestions broaden my vision and guided me to succeed in this
work.
I extend my thanks to Dr. G. S. Gawande, Head of Electronics &
Telecommunication Engg. Department, Shri Sant Gajanan Maharaj College of
Engineering, Shegaon for their valuable support that made me consistent performer.
I also extend my thanks to Dr. S. B. Somani, Principal, Shri Sant Gajanan
Maharaj College of Engineering, Shegaon for their valuable support.
Also I would like to thanks to all teaching and non-teaching staff of the
department for their encouragement, cooperation and help. My greatest thanks are to
all who wished me success especially my parents, my friends whose support and care
makes me stay on earth.
Place: SSGMCE, Shegaon Nidhi V. Shirbhayye

5. v
SSGMCE, Shegaon
Contents
Abstract iii
Acknowledgement iv
Contents v
List of Figures& Tables vi
Abbreviations vi
1. Introduction 1
1.1 Definition 2
1.1.1 Topic Model 2
1.1.2 Latent Dirichlet Allocation 3
1.1.3 Latent Motiom Patterns 3
1.1.4 Optical Flow 4
1.1.5 Log-likelihood 4
2. Literature Review 5
2.1 Magnetic Loop Detector 5
2.2 Smart Video Survellience System For Vehicle Detection and Traffic 6
Flow Control.
2.3 A real-time computer vision system for vehicle tracking and traffic 7
surveillance
2.4 Topic model Approach 8
3. Background Theory 9
4. Methodology 11
4.1 ROI Determination 11
4.2 Video Representation 12
4.3 Model Learning 13
4.4 Traffic Density estimation 14
5. Conclusion 15
References 16

6. vi
SSGMCE, Shegaon
List of Figures and Tables
Figure 1.1 –Example of Motion Patterns.
Figure 2.1 –Magnetic Loop Detection.
Figure 3.1 – Graphical model representation of LDA.
Figure 4.1 – Block diagram of a traffic density estimation based
on Topic Model
Figure 4.2 – Traffic scene
Figure 4.3 – ROI
Figure 4.4 –Preprocessing of Image and Video Data
Figure 4.5 – Extracted Optical Flow
Abbreviations
HMM -Hierarchical Hidden Markov Model
RFID -Radio Frequency Identification
ITS -Intelligent Transportation System
LDA -Latent Dirichlet Allocation
ROI -Region of Interest
SVSS -Smart Video Surveillance software
UCSD -University of California San Diego

7. vii
SSGMCE, Shegaon
1 Introduction
Increase in traffic is one of the major concerns for the city development. The
number of vehicles on the road increases day by day therefore for the best utilization
of existing road capacity, it is important to manage the traffic flow efficiently. Traffic
congestion has become a serious issue especially in the modern cities. The main
reason is the increase in the population of the large cities that subsequently raise
vehicular travel, which creates congestion problem.
Due to traffic congestions there is also an increasing cost of transportation
because of wastage of time and extra fuel consumption. Traffic jams also create many
other critical issues and problems which directly affect the human routine lives and
some time reason for life loss.
Under this circumstance, the conventional traffic light systems which are timer
based are not able to control traffic congestion. If a lane has more traffic congestion
than the others, the existing system fails to control traffic.
To solve this problem, a real time traffic control system is needed which will
control the traffic light according to traffic density. The conventional traffic system
needs to be upgraded to solve the severe traffic congestion, alleviate transportation
troubles, reduce traffic volume and waiting time, minimize overall travel time,
optimize cars safety and efficiency, and expand the benefits in health, economic, and
environmental sectors.
Road traffic density estimation provides important information in Intelligent
Transportation Systems (ITS). A real time area based traffic density estimation
method which will help an intelligent traffic control system to control traffic light
according to traffic density. Accurate calculation of traffic density is essential for the
development of early warning and automatic signaling Systems. Moreover, density
data can be used to help drivers for choosing optimal way among variety of routes [1].

8. viii
SSGMCE, Shegaon
There are lots of techniques proposed to design an intelligent traffic system, for
example, fuzzy based controller and morphological edge detection technique are
proposed. This technique is based on the measurement of the traffic density by
correlating the live traffic image with a reference image. The higher the difference is,
higher traffic density is detected. In some technique of controlling the traffic signal
by using image processing, in which first the reference image is selected, which is the
image with no vehicles or less vehicles and every time matching real time images
with that reference image. On the basis of the percentage of matching traffic lights
controlled. But in this technique image matching is performed by the edge detection.
The reference subtraction is a complex technique, with limited outcomes.
A new method for traffic density estimation is proposed, which provides us
more accurate information for signal decision making. Density forecasting is done by
extracting low-level features and applying topic models.
1.1 Definition:
1.1.1 Topic Model:
Topic models were first introduced to discover latent topics in a large collection of
textual documents. In machine learning and natural language processing, a topic
model is a type of statistical model for discovering the abstract "topics" that occur in a
collection of documents.
Topic modeling is a frequently used text-mining tool for discovery of hidden
semantic structures in a text body. Intuitively, given that a document is about a
particular topic, one would expect particular words to appear in the document more or
less frequently.Using the concept of Topic Model, approach of estimation of traffic
density is is explained in this seminar report. In this approach LDA is used.

9. ix
SSGMCE, Shegaon
1.1.2 Latent Dirichlet Allocation (LDA):
It is a generative probabilistic model which enforces a Dirichlet prior over the
topic distributions and word distributions. LDA is a generative probabilistic model for
collections of discrete data such as text corpora[4]. This special characteristic makes it
the most popular topic model.
In natural language processing, Latent Dirichlet Allocation (LDA) is
a generative statistical model that allows sets of observations to be explained
by unobserved groups that explain why some parts of the data are similar.
For example, if observations are words collected into documents, it posits that each
document is a mixture of a small number of topics and that each word's creation is
attributable to one of the document's topics.
1.1.3 Latent Motion Patterns:
The obtained direction and magnitude models learn the dominant motion
orientations and magnitudes at each spatial location of the scene and are used to detect
the major motion patterns. In many surveillance scenarios, such as monitoring traffic
at intersections, crowded video scenes with various motions may be involved. In these
scenes, some typical activities, called motion patterns, occur regularly and
periodically.
It is highly desired to analyze the motion patterns and extract some type of
high-level interpretation of the video contents. Discovering such motion patterns
would directly lead to a semantic scene model that could further facilitate the task of
scene analysis. Analyzing motion patterns in traffic videos can directly lead to
generate some high-level descriptions of the video content.

10. x
SSGMCE, Shegaon
Fig. 1.1 Example of motion patterns
1.1.4 Optical Flow:
Motion estimation generally known as optical or optic flow. Lucas kanade
method is one of the methods for optical flow measurement. It is a differential method
for optical flow estimation.Optical flow or optic flow is the pattern of apparent motion
of objects, edges and surface in a visual scene caused by the relative motion between
an observer (an eye or a camera) and the scene[6].The corner detector is utilize to find
the key points and use this features to extract the optical flow using Lucas–Kanade
method from each pair of consecutive frames.
1.1.5 Log-likelihood:
In statistics, a likelihood function (often simply the likelihood) is a function of
the parameters of a statistical model given data. Likelihood functions play a key role
in statistical inference, especially methods of estimating a parameter from a set of
statistics.
In informal contexts, "likelihood" is often used as a synonym for "probability." In
statistics, a distinction is made depending on the roles of outcomes vs. parameters.
Probability is used before data are available to describe possible future outcomes
given a fixed value for the parameter (or parameter vector). Likelihood is used after
data are available to describe a function of a parameter (or parameter vector) for a
given outcome.

11. xi
SSGMCE, Shegaon
2. Literature Review
In the past, road engineers tried to estimate the traffic flow or traffic density on a road
by using magnetic loop detectors or supersonic wave detectors, some operators
manually estimate the traffic density that was so boring and inefficient.
2.1 Magnetic Loop Detector:
Vehicle detection loops, called inductive loop or Magnetic Loop traffic detectors, can
detect vehicles passing or arriving at a certain point, for instance approaching a traffic
light or in motorway traffic. An insulated, electrically conducting loop is installed in
the pavement. The electronics unit transmits energy into the wire loops at frequencies
between 10 kHz to 200 kHz, depending on the model. The inductive loop system
behaves as a tuned electrical circuit in which the loop wire and leadin cable are the
inductive elements. When a vehicle passes over the loop or is stopped within the loop,
the vehicle induces eddy currents in the wire loops, which decrease their inductance.
The decreased inductance actuates the electronics unit output relay or solidstate
optically isolated output, which sends a pulse to the traffic signal controller signifying
the passage or presence of a vehicle[8].
But now, traffic management systems utilize image and video processing techniques
to extract the same information from videos captured of different roads or junctions.
Video monitoring and automatic traffic flow detection, as the most commonly used
method in traffic management, takes precedent over the traditional methods. It can
provide highquality image information efficiently and stably, without damaging the
road or blocking the traffic.

12. xii
SSGMCE, Shegaon
Fig. 2.1 Magnetic Loop Detection
2.2 Smart Video Survellience System For Vehicle Detection and
Traffic Flow Control
Author: A. A. Shafie, et al.
Traffic signal light can be optimized using vehicle flow statistics obtained by
Smart Video Surveillance Software (SVSS). This research focuses on efficient traffic
control system by detecting and counting the vehicle numbers at various times and
locations. At present, one of the biggest problems in the main city in any country is
the traffic jam during office hour and office break hour. Sometimes it can be seen that
the traffic signal green light is still ON even though there is no vehicle coming.
Similarly, it is also observed that long queues of vehicles are waiting even though the
road is empty due to traffic signal light selection without proper investigation on
vehicle flow. This can be handled by adjusting the vehicle passing time implementing
by this developed SVSS. A number of experiment results of vehicle flows are

13. xiii
SSGMCE, Shegaon
discussed in this research graphically in order to test the feasibility of the developed
system. Finally, adoptive background model is proposed in SVSS in order to
successfully detect target objects such as motor bike, car, bus, etc.
The advent of computer vision and digital image processing technology and its
development considerably help video based traffic flow detection system to become
increasingly robust, real time and intelligent. Due to these advantages, video based
traffic management and surveillance systems are becoming more and more significant
to ITS.
2.3 A real-time computer vision system for vehicle tracking and
traffic surveillance
Author: Benjamin Coifman, et al.
Increasing congestion on freeways and problems associated with existing detectors
have spawned an interest in new vehicle detection technologies such as video image
processing. Existing commercial image processing systems work well in free-flowing
traffic, but the systems have difficulties with congestion, shadows and lighting
transitions. These problems stem from vehicles partially occluding one another and
the fact that vehicles appear differently under various lighting conditions. Scientists
are developing a feature-based tracking system for detecting vehicles under these
challenging conditions. Instead of tracking entire vehicles, vehicle features are tracked
to make the system robust to partial occlusion.
The system is fully functional under changing lighting conditions because the
most salient features at the given moment are tracked. After the features exit the
tracking region, they are grouped into discrete vehicles using a common motion
constraint. The groups represent individual vehicle trajectories which can be used to
measure traditional traffic parameters as well as new metrics suitable for improved
automated surveillance. This method describes the issues associated with feature

14. xiv
SSGMCE, Shegaon
based tracking, presents the real-time implementation of a prototype system, and the
performance of the system on a large data set.
There are numerous researches completed on vehicle counting and traffic density
estimation in the recent literature.
 Arora and Banga used fuzzy based controller and morphological edge
detection method. Their approach found traffic density by computing
correlation between live traffic image and an initial image, in which higher
difference demonstrated the higher traffic density.
 Dangi proposed another method based on four lanes system, in which the
number of vehicles on the lane determined the allocated time.
 The approach proposed by Gupta et al. computed the traffic load in each live
image by comparing it and the reference image.
 Kanojia suggested another method based on image processing to control the
traffic signals. He first selected the reference image that have no vehicles or a
few number of vehicles, and matched the given images with that reference
image by the edge detection. Then, traffic lights were controlled based on the
percentage of this matching.
 Clarkson present an approach for unsupervised decomposition of on-body
sensor data into events and scenes. They use data from wearable sensors to
discover short events such as ”passing through a door” or ”walking down an
aisle”, and cluster these into scenes such as ”visiting the supermarket” by
using hierarchies of HMMs.
 Suresh Sharma had discussed about RFID. the use of RFID traffic control to
avoid problems that usually arise with standard traffic control systems,
especially those related to image processing and beam interruption techniques
are discussed. This RFID technique deals with multivehicle, multilane, multi
road junction areas. It provides an efficient time management scheme, in
which, a dynamic time schedule is worked out in real time for the passage of
each traffic column.
Hence, after going through these literatures, a new framework for traffic density
estimation based on topic model, which is an unsupervised model is putforth in this

15. xv
SSGMCE, Shegaon
seminar report. This framework uses a set of visual features without any need to
individual vehicle detection and tracking, and discovers the motion patterns
automatically in traffic scenes by using topic model. Then, likelihood value allocated
to each video clip enables us to estimate its traffic density. Results on a standard
dataset show high classification performance of our proposed approach and
robustness to typical environmental and illumination conditions.
2.5 Topic Model Approach:
In topic model, users’ activities are seen as a collection of “documents”, and the
components of activities are “words”. Each document is a collection of words. The
process of training a classifier to recognize complex activities based on multiple
devices is done. In the same way topic model can be applied to video data for cunting
number of vehicles.
First, the video is taken and region of interest is separated out for feature extraction of
frames of particular length from it. Then LDA is applied to that video data. A key
factor needs to be considered is the length of windows. If the windows are too short,
they may not provide sufficient information to describe a complete number of
vehicles. Quantization of the position and direction and magnitude of optical flow
vectors is done for generating the visual words. Topic model is applied to discover
the set of latent motion patterns from video by learning the distribution of visual
features that occur at the same time, and the distributions of motion patterns that
cooccur in the video is learnt from it.

16. xvi
SSGMCE, Shegaon
3. Background Theory
Topic models were first introduced to discover latent topics in a large collection of
textual documents . Then some researchers have utilized these models for video
analysis . One of the most common and successful topic models is Latent Dirichlet
Allocation (LDA).
It is a generative probabilistic model which enforces a Dirichlet prior over the
topic distributions and word distributions. This special characteristic makes it the
most popular topic model and briefly reviews in the following:
In the collection of Nd documents, if we assume there are K topics, each
document d is modeled as a mixture of these K topics that is shown as Ѳd. In
addition, each topic k is modeled as a multinomial distribution over a vocabulary
given by β={βk}. For each document d, a parameter Ѳd of the multinomial
distribution is drawn from Dirichlet distribution Dir(Ѳd,α), which α is a Dirichlet
prior on the documents. For each word wdn in document d, a topic zdn is drawn with
probability Ѳdk, and word wdn is drawn from a multinomial distribution given by
β(zdn). α and β are the hyperparameters that must be optimized to get optimal topics.
Given the parameters α and β, then the joint distribution of a topic mixture Ѳ, a set of
N topics z, and a set of N words w is expressed by:
P (Ѳd , zd ,wd |α ,β ) = P (θd | α) ∏ 𝑃( 𝑧𝑑𝑛, Ѳ𝑑) 𝑃(𝑤𝑑𝑛⃒𝑧𝑑𝑛, 𝛽)𝑁
𝑛=1 (1)
Since the marginal likelihood P(wn|α, β) and the posterior distribution
P(Ѳd,zd|α,β) are intractable for determining exact inference, an inference method,
such as variational Bayesian (VB) method is utilized to approximate P(Ѳd,zd|α,β) .

17. xvii
SSGMCE, Shegaon
Fig. 3.1 Graphical model representation of LDA. The boxes are “plates” representing
replicates. The outer plate represents documents, while the inner plate represents the repeated
choice of topics and words within a document.

18. xviii
SSGMCE, Shegaon
4. Methodology
The traffic density estimation system includes four main components: ROI
determination, feature extraction and construction histogram of words, model
learning, and density estimation. The block diagram of the present paper is illustrated
in Fig. 4.1. The details of these components are explained in following subsections.
Fig. 4.1 Block diagram of a traffic density estimation based on Topic Model
4.1 ROI Determination:
The first step is to select region of interest (ROI) where the vehicle of interest road
lane are present. The purpose of selecting ROI is to exclude the unnecessary
background information such as other road lane. Since the camera is stationary, this
unnecessary information is fixed in frames of the live video.

19. xix
SSGMCE, Shegaon
Fig. 4.2 Traffic scene Fig. 4.3 ROI
4.2 Video Representation:
The second step is the feature extraction in order to represent the video by features
and construct the histograms of words.
 To perform the feature extraction, we first temporally divide the entire video
into Nd non-overlapping short clips. Then corner detector is employed to extract
the key points.
 For each pair of consecutive frames, key points are used to discover the optical
flows using Lucas-Kanade method.
 In order to remove noise and preserve only the reliable flows, we apply a
threshold TH0 to the magnitude of optical flow vectors. For generating the
visual words, the position and direction and magnitude of optical flow vectors
are quantized.
Fig. 4.4 Preprocessing of image and video data

20. xx
SSGMCE, Shegaon
Fig. 4. 5 Extracted Optical flow
 After spatial, directional and magnitude quantization, a vocabulary of visual
words is obtained, in which each word present three aspects of contents, such as
information about position, motion direction, and velocity of motion.
Optical flow vectors are denoted by (x, y, α, λ). The position (x, y) are quantized
to the nearest position on a grid with spacing of H pixels and the angles of flow
vectors, α, are quantized into Nm directions. Also, the magnitudes of flow vectors, λ,
are quantized into Nc values. A vocabulary V of N=Na×Nb×Nm×Nc visual words:
V={vi},i=1,…N, is obtained. Histogram of words which constitute the inputs of the
topic model are created by means of accumulating the visual words over the frames of
each video short clip. Then, a clip dj of video D={di} i=1,…Nd, is represented as a
vector W={wn} n=1,…N, where wn denotes the number of occurrence of word n in
the clip. Then, D is given to the topic models in order to correlations among these
visual words.
4.3 Model Learning:
The third step is model learning to learn the motion patterns in video clips. Topic
model is used to discover the set of latent motion patterns from video by learning the
distribution of visual features that occur at the same time, and to learn distributions of

21. xxi
SSGMCE, Shegaon
motion patterns that cooccur in the video. Then, these learned motion patterns are
employed to calculate likelihood measure to estimate traffic density in traffic videos.
4.4 Traffic Density Estimation:
In order to calculate the traffic density, the topic model is trained with specific
density such as light-density first and then estimate the traffic density by using log-
likelihood measure at the end of the fitting phase. Thus, the clips with the same
density as the training dataset will produce high log-likelihood. On the contrary, the
clips which contain different density with training dataset will achieve low likelihood,
because learned topics is not able to describe the observed visual words of that
density. Since the likelihood is not normalized, this measure is highly dependent on
the clip size. Therefore, to overcome this issue, we divide log-likelihood of each clip
by the number of visual words in that clip and name it normalized log-likelihood.
Also, we use two thresholds to determine type of the traffic density such as light-
density, medium-density and high-density.

22. xxii
SSGMCE, Shegaon
5. Conclusion
By comparing with other works which had performed traffic density estimation.
The classification accuracies of these works were 93.3% to 95.3%[1]. A drawback of
these approaches is high computational cost of fitting models that makes them
impractical for application to real-time traffic monitoring. Therefore, although our
classification accuracy is lower than that of them, our framework can be employed in
real-time applications.
This report presented a framework to automatically classify complex traffic videos
and determine their traffic density, based on LDA, which is one of the most successful
topic models. Results on UCSD database showed that this framework is able to
accurately estimate the density of traffic videos even in bad illumination condition[1].
The overall classification accuracy of our proposed framework was achieved equal to
92.2%[1]. Moreover, since the execution time for our approach is relatively low, it
can be used in real-time application. In the future works, its results would have been
corroborated by more datasets.

23. xxiii
SSGMCE, Shegaon
References
[1] Razie Kaviani, Parvin Ahmadi, Razie Kaviani, Parvin Ahmadi,”A New Method for Traffic
Density Estimation based on Topic Model”, SPIS2015, 16-17 Dec. 2015, Amirkabir University of
Technology, Tehran, IRAN ,IEEE.
[2] A. A. Shafie, M.H. Ali, Fadhlan Hafiz, roslizar M. Ali “Smart video surveillance system for
vehicle Detection and traffic flow control” Journal of Engineering Science and Technology
Vol. 6, No. 4 (2011) 469 - 480
[3] Mohammad Shahab Uddin, Ayon Kumar Das, and Md. Abu Taleb, “Real-time area based traffic
density estimation by image processing for traffic signal control system: Bangladesh
perspective”, in 2nd Int’l conf. on Electrical Engineering and Information and Communication
Technology (ICEEICT) 2015,IEEE, 21-23 May 2015.
[4] David M. Blei, Andrew Y. Ng., Michael I. Jordan, ”Latent Dirichelet Allocation”, Journal of
Machine Learning Research 3 (2003) 993-1022.
[5] Parvin Ahmadi, Soroosh Khoram, Mohsen Joneidi, Iman Gholampour, Mahmoud Tabandeh, ”
Discovering Motion Patterns in Traffic Videos using Improved Group Sparse Topical Coding”,
2014 7th International Symposium on Telecommunications (IST'2014), IEEE.
[6] Dhara Patel, Sourabh Upadhyay, “Optical Flow Measrement Using Lukas kanade Method ”,
International Journal of Computer Applications (0975 – 8887) Volume 61– No.10, January 2013.
[7] Benjamin Coifmana, David Beymerb, Philip McLauchlanb,Jitendra Malikb, “A real-time
computer vision systemfor vehicle tracking and traffic surveillance”
[8] Retrieved from Likelihood function at Planetmath (http://planetmath.org/likelihoodfunction).
[9] Retrieved from external link of Traffic sensor
(http://auto.howstuffworks.com/cardrivingsafety/safetyregulatorydevices/question234.htm) from
How Stuff Works.

24. xxiv
SSGMCE, Shegaon