BestKet - Kaan Demirel, Ufuk Bugday

Introduction

Project Definition

A free throw is a deciding factor in basketball. It can bring you the championship trophy or can send you home early. If your team is surrounded with good shooters, then you might not worry about missing free throws that can cost you the game, but if you have some bad free throw shooters, that can end up changing your whole strategy for the game. With our system, we are aiming to make the players more successful at the free throw line which will prevent the teams from changing their offensive strategies.

Project Plan

Work Breakdown Schedule

WBS will show our project plan with way things are going to work in. With this you can see the evaluation and the phases of work. Every phases will be detailed in our gantt chart. We will show the importance of the levels on other phases

Project Proposal

With our project, we are aiming to improve the free throw percentage of any player depending on some basic factors.

Field Research

As far as we know there are no programming project about that but there are some studies about probable best shot formula.

Proposal

Enhancing %20 of success of any player who had below %50 is our main proposal.

Project Plan

Preparing the Project Plan

We prepared this project plan document to be able to move forward easily and planned.

Analysis

Assumptions

-High definition free throw footage is required. (min 720p)

-The footage should have no other basketballs or players.

-The following footage angles are necessary:

-Left (Example Footage)

-Right

-Back

-Left and Right Shooter Close-up

Determining Requirements

After doing some research, we came up with these factors:

1) Before the shot, determine the following distances and angles from the video stream:

-Left Foot Position(Distance, Angle)

-Right Foot Position(Distance, Angle)

-Shooter(Length, Standing Reach)

-ShootingArmAngle(Ground,UpDown)

Ground is angle number 2

UpDown is angle number 1

-TheAngleBetweenFootAndLowerLeg

Angle number 4

-TheAngleBetweenLowerAndUpperLeg

Angle number 3

2) On release:

-Ball Position(Distance to Hoop, Release Velocity)

-Left Foot Position(Distance, Angle, HeelStatus)

-Right Foot Position(Distance, Angle, HeelStatus)

-ShootingArmAngle(Ground,UpDown)

Ground is angle number 2

UpDown is angle numer 1

-NumberOfShootingFingers

-TheAngleBetweenFootAndLowerLeg

Angle number 4

-TheAngleBetweenLowerAndUpperLeg

Angle number 3

Design and Programming

UI Design

For the UI, we are thinking about a simple screen which requests video streams, and returns the suggested factors for the player on the video.

Image Processing Implementation (Collecting Data)

To collect data, we selected to use OpenCV library. We are currently exploring different algorithms that can do what we are expecting.

For ball detection, we will use OpenCV's findContours function.

For human joints detection and stick figure implementation, we will be choosing between Temporal Sequential Forests Approach or Zheng-Suon Thinning Algorithm depending on their effectiveness.

Best Shot Formula Implementation

Still researching the impacts of the factors...

Testing

Will be updated later on.

Estimates

We are aiming to finish the project by May, 2016. Detailed estimations on tasks can be found below:

   Project Proposal (September – End of September)

   Project Plan (End of September – End of October) 

   Analysis (Mid of October – Beginning of December)

   Design and Programming (Beginning of December – End of March)

   Testing (Beginning of February – End of May)

Resources

Software Development Process

We decided to develop our program with using Waterfall Method as a sequential design process because it is easy to understand. Also, to proceed to next part, each part must be completed fully. At the end of each section we have to determine if the project is on the right way.

Schedule

Logs

Week 1&2

We try to decide what we can do as a senior project. Since we are basketball fan we decided to work on one of the major basketball problem which is improvement of free throw percentage. To do that we outlined the basics of project then we decided that we could handle this via OpenCV with video processing. After we specified waterfall method as a sequential design process. We needed to some data for specifying factors.

Week 3&4

We went to gym to record some free throw shots. We needed some restrictions to record a clip and these were a camera with at least 720p video quality, no other basketball or player on the background (just one shooter and one ball), stable and tree different angle which are left, right and back. After some records we noticed what factor we should take care of for the calculating best free throw shot. For example distance between player and hoop, shooting elbow angle, .etc.

Week 5&6

We did some research for image processing algorithms. Searched for the best way to track the basketball. Tried implementing Hough Circles for ball detection but it gave lots of false positives. For the tracking ball we found out "Magnus Effect" and Stoke's Law for the best way for tracking ball. The Magnus Effect is a force that give spin to the ball. It is obvious that player could not decide the magnitude of spin but we are trying to calculate trajectories for the ball. We will do this probably using 2D model rather than 3D model so that we can ignore the 3rd dimension errors. The air resistance is the another factor to considerate. We found out some formula which are driven from Stoke's Law may be useful to formulate the best shot formula.

Fd = 6πµRv0

Fd - the frictional force µ - dynamic viscosity (can be constant for some particular objects in case we are looking for basketball) R - radius of basketball v0 - intial calculated velocity

For future we probably will use this equation to restrict our best shot formula.

Week 7&8

We chose to go with findContours for the ball detection, because we find out that findContours are much more stable than Hough Circles when it comes to circle detection on an object with a specific color. For stick figure implementation we found two different approaches: Temporal Sequential Forests Approach(implemented only for upper body) and Zheng-Suon Thinning Algorithm(no detection algorithm but eases the way for human joints detection) In spite of deciding which algorithms we are going to use, we worked on our UI design. Decided that our UI will require a screen for users to upload an example free throw shooting video. After processing the video, it will display the factors of the video and the factors that shooter needs for a perfect shot. Users will also have an option to see an example free throw shot animation if they want to. To create the animation we are thinking about using Unity Engine. We are still researching for other animation creating options.