He Shoots! He Scores! Here’s How!

UBC Reports | Vol. 49 | No. 11 | Nov.
6, 2003

By Michelle Cook

In the fast-paced world of hockey, there isn’t a coach
— or fan — alive who hasn’t at one time wished for
more insight into why some plays unfold perfectly while others
end in disaster.

Now, thanks to a UBC computer scientist who can’t
even skate, coaches may soon have a sophisticated new tool
to help them analyse and predict how players will perform
during a game, or even choose their top draft picks.

Before Kenji Okuma, 25, came to Canada from Japan in July
2000, he’d never seen a hockey game or set foot on a
rink. Three years later, he’s watched hundreds of hours
of NHL action on videotape — paying special attention to
goal highlights — in order to build an unusual database.

Okuma is part of a team of UBC researchers working to develop
a computer system capable of plotting player movements in
hockey games. Its members include computer vision specialists
James Little, David Lowe and Robert Woodham, and data mining
expert Raymond Ng — all professors in the computer science
department. Okuma completed his MSc in computer science earlier
this year, and has been working as a research assistant in
the department’s Lab for Computational Intelligence
since then.

The research team’s goal is to create a large database
of NHL players in motion that can be queried to extract the
paths — or motion trajectories — of individual players.
The patterns could then be analysed to determine how a player
moves and how he would be likely to move in future plays.
Similar systems already exist for baseball, soccer, football
and tennis.

Woodham says that hockey was the sport of choice for the
project, in part, because of the interest in it here in B.C.
and also because he and fellow researcher Little are die-hard
fans of the game.

Nonetheless, Okuma says he was drafted for his off-ice skills.

“Jim (Little) was my thesis advisor. I told him I knew
nothing about hockey,” Okuma says. “I’d
never played hockey but I was living in Canada and thought
I should know what Canada’s favourite sport was like.”

Okuma quickly learned what avid fans already know: professional
hockey is a fast sport, and keeping an eye on the puck can
be hard work. When a game is broadcast, cameras tilt, zoom
and pan and switch between different locations, but they only
capture a side or end view of the rink. To accurately understand
the way players move around the ice, the best viewpoint for
a coach is looking down on the rink from above.

With little overhead game footage available to the researchers,
Okuma created software that removes the camera motion, and
isolates and tracks each player’s route. The software
then transforms regular broadcast video footage into a digitized
top-down view of the players’ movements that is useful
for coaches and other analysts.

“The human eye can only track three people at a time
so this data would give a coach a fuller picture of where
and how most plays occurred in a game,” Okuma says.

He adds that the research team would like to get access
to more game footage than is currently available on television
broadcasts. An increased number of camera angles around hockey
rinks would allow them to capture a greater range of movement
on the ice, and help them to more accurately replicate human
motion in digital form. The researchers would also like to
get professional coaches interested in trying out the system.

Hockey coaches and sports analysts are the most obvious
end-users of the research, which is being funded by the Institute
for Robotics and Intelligent Systems (IRIS) as part of a larger
nationwide project of intentional motion and data archiving,
but the work being conducted has wider applications.

Sports broadcasters could use it to give TV viewers a more
dynamic experience by providing new vantage points and analysis
of game play. Off the rink, the system could aid the computer
game and film industries to create more realistic animated
characters, or be used in the development of smart robots
capable of imitating human motion or anticipating people’s
actions.

With his contribution to the project almost complete, Okuma
will be heading back to Tokyo in December with plans to pursue
a PhD. And, he says, after watching two full seasons of NHL
action on tape, he’s had his fill of TSN broadcasts
and Hockey Night in Canada and will be happy to get back to
his sport of choice — soccer.

Still, he thinks he might like to go down to GM Place just
once before he leaves. Despite the hundreds of games he’s
viewed, he’s never seen a game of hockey played live.

-

-

-