“Intelligent Transportation Systems (ITS)” is a phrase that is being given great emphasis in India,

today. Metropolitan cities like Bangalore have introduced, not just high-tech public transportation

buses, but also certain speed detection and speed management systems. However, These systems though,

were designed under the pretext that the road traffic is characterised as organised, homogenous,

and one that abides with lane discipline.

As a student of the Department of Civil Engineering, at the National Institute of Technology –

Tiruchirappalli, my final year project for the past 4 months was -–is todevelop an efficient algorithm for vehicle

movement detection, to then detect the speed of the vehicle and propose a framework for speed

management. From the start, my project took an interdisciplinary approach. Concepts of image

classification such as an RGB image, a Greyscale image and a Binary image were engrossing. [AM1]In

general, the primary concepts of image processing were to be understood and assimilated, in

order to proceed. This then led me to the next step, which was to review and comprehend various

published papers in the field of speed detection techniques which that employ image processing. At this

juncture, I faced an exciting challenge - in trying to understand the new methodologies that were

being suggested to achieve similar goals. Some of them incorporated mathematical theories such as

Fourier Transforms theory[AM2].When I could [AM3]redefine the problem statement of the project to myself, I

was able to address issues such as ‘feasibility’ and ‘clarity’.

The framework for speed management was to be developed for Tiruchirappalli,(a city in Tamil

Nadu) and the location of my campus. I studied the traffic patterns of 24 existing roads in the city for 30 days, and after generating

graphical speed profiles at peak and non-peak hours of each day per road, I could narrow the

number down to 6 roads – ones that need to be studied further in detail.[AM4]

After analysing each of those 6 roads, I chose the one with the most dynamic and complex

background[AM5] (alternate suggestion: After analysing each of the 6 roads and by scoring them on various categories, such as volume, peak velocity, traffic-time profile; I then selected roads in manner that gave me the best diversity in the data). This selection was based on Phase 1 – the methodology of detecting the moving

vehicle. [AM6]A minor challenge that we faced here was in selection of the appropriate camera angle

– whether it should be an overhead view, a side view, or a front angled view of the vehicle. The

answer was rather exciting [AM7]because we[AM8] needed to choose a view which had a complex background

to account for reality. Thus the side view of the road was chosen to be the best angle. [AM9]The

background image was obtained devoid of any dynamic foreground objects. Similarly, we[AM10] took a

video of a vehicle travelling across with the same background image.

Hence, in a single frame with the moving object I subtracted the background to reveal the

foreground feature. The most challenging and the most prominent part here was when I performed

certain image processing operations in a widely known software called MATLAB[AM11], using an

inbuilt ‘OpenCV library’ [AM12]of functions, on the final image to obtain the centroid of the vehicle in side

view, by creating a bounding box – a rectangular box covering the extremities of the vehicle in side

view.

Amongst the few image processing operations I employed, ‘Thresholding’ and ‘Image smoothing’

were two prominent ones. Their functions and working processes on the image were most exciting[AM13]

to me in a form that they add meaning to my final processed image. The former refers to the process

of filtering out pixel values in an image by applying a threshold value, within a fixed frequency

domain. On the other hand, the latter refers to the process of eliminating unwanted noise from

the image. This is achieved by employing a certain ‘image processing blur’ operation often called

as ‘Gaussian Blur’, which uses the Gaussian function to blur the image. [AM14]It uses a simple concept

which states that – the greater the factor to which an image is shrunk to, causes a greater degree of

blur when resized back to the original.[AM15]

Phase 2 was to develop a ‘Methodology for Speed Detection’ of the moving vehicle. The two most

critical points that I had to take into account were - the accuracy of the centroid values taken which

depend on the degree to which noise removal was optimal, and the number of available frames per

second (i.e., the frame frequency) in the video.

At first, I calibrated the camera and the program code for the particular location by knowing the

distance between the two horizontal extremities. Further, I divided this distance with the width of

the video frame (in pixels) to obtain the ‘distance to per pixel’ factor. Finally, the distance moved by the

centroid of the vehicle between two frames was calculated and divided by the time per frame, to

obtain the velocity of the subject vehicle.

I believe the sole reason that this project appealed to me was that it involved a wide spectrum of

possibilities, theories, new concepts and tasks that kept me agile [AM16]and alert at all times till fruition.

Its interdisciplinary nature allowed me to have a raw view of the contents of image and video

processing, along with an understanding of ‘Intelligent Transportation Systems’ as a vital component

of‘Traffic Engineering’[AM17]. Through the course of the project I have realised that possessing a multi

dimensional approach can provide a myriad of pathways and viewpoints to counter constraints

in order to achieve a solution. [AM18]Thus, I am of the opinion that the Career Discovery program at

Harvard University will inculcate in me a capacity to view cities with more than one perspective as

a responsive Urban Planner. As a result, I would be a well informed student of Urban Planning and

Urban Design and thus give cities a much-demanded sustainable outlook in the future in their plan

and design.

[AM1]Sounds much too noob.

[AM2]What are you talking about ?

[AM3]What are you trying to say here ?

[AM4]When the hell did you do all this ??

[AM5]Big words, no meaning.

[AM6]How is the selection based on the motion detection?

[AM7]How is the answer exciting ?

[AM8]Maintain either an ‘I’ stance or a ‘We’ stance. Or you can say “my team”

[AM9]We chose the side angle owing to feasibility and the fact that once we can have some basic video feed, we can work on the general detection algorithm.

[AM10]Again the ‘we’ – ‘I’ clash.

[AM11]They should know what MATLAB is I guess, You don’t have to introduce it like it’s new.

[AM12]WHAT ??

[AM13]You use ‘exciting’ a bit too much.

[AM14]Very loose. Tighten it up.

[AM15]Dude, this is not Gaussian blur.

[AM16]How were you ‘agile’ because of a software project?

[AM17]Why is this in quotes?

[AM18]Sounds much too roundabout. Just hit the points. You have a word limit.