C10 Vision Sciences Society2004

Is Multiple Object Tracking Color Blind??Sarah B. Klieger1, Todd S. Horowitz1,2, & Jeremy M. Wolfe1,2

1Brigham & Women’s Hospital and 2Harvard Medical School

V E R I TAS

Ave

rag

e C

orr

ect

Background: Features

In prior work, observers had difficulty reporting changes in featural properties of tracked objects (Scholl, Pylyshyn, & Franconeri, 2004). However, those features might still be used to individuate items and improve tracking.

Objects can usually be distinguished by various features, including, size, shape and color.

Experiments 2 and 3 investigate the effects of a color feature on tracking. Can color improve tracking by individuating items or by disinguishing a group of targets from distractors?

Background: Occlusion Depth cues improve tracking when items overlap (Viswanathan & Mingolla, 2002), and occluding surfaces do not impair tracking (Scholl & Pylyshyn, 1999).

Therefore, it has been widely assumed that allowing items to occlude one another should not impair tracking. However, Yantis’ (1992) data suggest otherwise. In Experiment 1, we directly tested this prediction.

We used two types of displays. In one, items “bounced” off each other. In the other, items could occlude each other.

Multiple Object TrackingWe often need to track several independently moving objects in our environment.

In the lab, observers are able to track ~ 4 to 5 independently moving targets among identical distractors (Pylyshyn & Storm, 1988), but in the real world, targets are typically heterogenous, differing on a number of features.

Today’s question: Can distinctive feature information improve tracking?

References• Pylyshyn, Z.W., & Storm, R.W. (1988). Tracking multiple independent targets: Evidence for a parallel tracking mechanism. Spatial Vision, 3, 179 - 197.

• Scholl, B.J., Pylyshyn, Z.W., & Franconeri, S.L. (2004). The relationship between property-encoding and object-based attention: Evidence from multiple object tracking. Submitted.

• Scholl, B.J., & Pylyshyn, Z.W. (1999). Tracking multiple items through occlusion: Clues to visual objecthood. Cognitive Psychology, 38, 259 - 290.

• Viswanathan, L., & Mingolla, E. (2002). Dynamics of attention in depth: Evidence from multi-elment tracking. Perception, 31, 1415 - 1437.

• Yantis, S. (1992). Multielement visual tracking: Attention and perceptual organization. Cognitive Psychology, 24, 295 - 340.

This research was supported by NIH grant MH65576.

General Conclusions

The visual system can track objects behind occluders, but inter-item occlusion presents a special problem: Moving stimuli can be confused with one another.

Using color to individuate objects does not help. However, this is not because the system cannot use color information.

When color segregates targets from distractors, inter-item occlusion is no longer a problem. This suggests that the challenge of multiple object tracking is segregating targets from distractors rather than individuating objects.

Future DirectionsWhat other attributes can be used to form groups of targets and groups of distractors?

Are spatiotemporal properties (such as motion type) helpful?

If inter-item occlusion only occurs among objects in the same category (targets, distractors), will it still be a problem?

Is it easier or harder to track unique objects?

General Methods

Initially, we presented a static display of 10 disks; 5 were flashed to cue them as targets. Objects moved at a speed of 6°/sec for 5 seconds. When motion ceased, observers identified targets with a mouse click.

On & Off for 2 Secs Track for 5 Secs 5 Responses

Inter-item occlusion significantly impairs tracking, even when depth cues, such as borders, are present.

Discussion

We can use this impairment to test the ability of features to aid tracking.

Results

Occlusion Bounce3.50

3.75

4.00

4.25

4.50

4.75

5.00

Ave

rag

e C

orr

ect

1. Does inter-item occlusion impair tracking?

We tested the hypothesis that inter-item occlusion would not interfere with tracking.

BounceOcclusion

3. Is tracking “colorblind”?

It seems surprising that vivid color had no positive effect on tracking.

Can tracking use color at all?

In Experiment 3, we gave the tracked set one color and the untracked set another color.

As before, colors shifted continuously during each trial

We can easily reject the hypothesis that tracking is colorblind.

Segregating targets and distractors by color eliminates the inter-item occlusion impairment, and makes tracking easier overall.

DiscussionResults

Occlusion Bounce3.50

3.75

4.00

4.25

4.50

4.75

5.00SegregatedMixed

Ave

rag

e C

orr

ect

Note: Ts designate targets. They are for illustrative purposes and were not present in the actual experiment.

Seg

reg

ated

M

ixed

Occlusion Bounce

TT

T

TT

T

T

T

TT

T

T

T

T

T

T

T T

T

T

We tested the hypothesis that tracking cannot use color information at all.

Results

Occlusion Bounce3.50

3.75

4.00

4.25

4.50

4.75

5.00

Unique

Same

Discussion

Color is not a feature that is used to individuate objects when tracking.

Color did not improve tracking in either condition.

2. Can color overcome the occlusion deficit?Each disk had a unique color to reduce inter-item confusion.

Colors shifted continuously during each trial.

initial frame

1 2

34

5

6

78

10

final frame

tracking

1 2

34

5

6

78

910

1

2

3

4

5

6

7

8 9

10

1

2

3

4

5

6

7

8

9

10

1

2

3

4

5

6

7

8

9

10

We tested the hypothesis that individuating objects by color could improve tracking.

Occlusion Bounce

Sa

me

U

niq

ue

Ave

rag

e C

orr

ect

Note: Numbers on the disks are for illustrative purposes and were not present in the actual experiment.