Introduction

For the past month, I have been working on a side project: ‘A Hybrid, Multi-Sensor Tracking System for Triple Jump Performance Analysis’.

As far as I am aware, such a system has not yet been built for triple jump and, more broadly, for field events in athletics for athlete/coaches use. The goal is to provide insights into the athlete’s technique. And answer questions such as: Is the horizontal velocity at take-off maximised through phases? Is the centre of mass (C.O.M.) in the correct position during transitions between different phases?

Purpose & Intended Use

The system aims to support the athlete and coach during training by identifying areas of improvement and providing visual cues to help prevent the athlete from repeating incorrect movements, which could lead to injury concerns due to repeated poor mechanics. Thus with this information, the coach may decide there is a need to focus training on improving speed maintenance between phases, refining transitions between the hop and step phases. Or optimising contact times and C.O.M. positioning to maximise distance. If the project is successful, my next step is to test it at training and then apply the same approach to the long jump (LJ), high jump (HJ), pole vault (PV), and throws.

Starting Point

To develop the system I am using, computer vision (CV), classical dynamics, electronics (i.e., inertial measurement units), and knowledge of TJ biomechanics. I have some familiarity with the latter. Since last June, I have been rehabilitating from a grade IV knee injury and to my suprise since October I have been training again with horizontal jumpers while learning more technical details about triple jump hopping and stepping. Applying my programming skills to this project seems quite fitting.

So far, I have been experimenting with the OpenCV library.

Video 1: Christian Taylor 18.21m (+0.2m/s) Triple Jump - 2015 Beijing, China

In terms of computer vision, I am mostly self-taught. While I was a student working on a quantum information research project at a research laboratory, I had the opportunity to talk to a computer vision scientist who worked in the same building. That is when I first considered applying computer vision to the long jump. Back in high school, my coach had already suggested building something for triple jump, knowing I planned to study physics at university. At the time, I did not see much reason to pursue it beyond personal interest and I thought a long jump version would be better. However, not only did it turn out I was a better triple jumper than a long jumper, during training last month, I discovered that a triple jumper needed something like this. After nearly a decade since first learning about computer vision (and even longer procrastinating on this idea) I have finally begun working on this project.

As for the physics, it should be quite interesting.

Video 2: Jonathan Edwards’ 17.86m Triple Jump Finals (3rd Round) - 2002 Commonwealth Games, Manchester, England

Design Considerations

For the first hardware prototype, I have chosen consumer-grade components. The goal is to design a hardware system with integrated Inertial Measurement Units (IMUs) that is affordable, portable and as compact as possible, since it needs to be wearable.

For the real-time video processing software component, I am writing software using Python, OpenCV and setting up a Logitech C922 (or C920).

So, let’s see how it progresses.

And as always, I hope you enjoyed reading.