The multiRANSAC Algorithm And Its Application to Detect Planar Homographies

by:
M. Zuliani, C. Kenney and B. S. Manjunath Vision Research Lab Department of Electrical and Computer Engineering University of California, Santa Barbara

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 1/22

Motivation - I
Estimation of model parameters in presence of noise and outliers is a crucial task in image processing, computer vision, patter recognition. . .
2 1.5 1 0.5 0 0.5 1 1.5 2 4 x Inliers Outliers Unbiased estimate (RANSAC) Biased estimate (least squares)

Least squares produce biased estimates in presence of outliers Robust statistic can tolerate up to 50% outliers

1 y

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 2/22

Motivation - I
Estimation of model parameters in presence of noise and outliers is a crucial task in image processing, computer vision, patter recognition. . .
Affine model 1

0.5

Outliers Inliers

RANSAC algorithm [Fischler 1981] RANdom SAmple Consensus

0.5

1 1.5 1 0.5 0 y 0.5 1 1.5

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 2/22

Motivation - II
What happens if there are multiple models?

0.8 0.6 y 0.4 0.2

0.2

0.4 x

0.6

0.8

(a)
(a) Multiple segments, (b) Multiple planes

(b)

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - II
Traditional RANSAC-based approaches: 1. apply standard RANSAC 2. remove the detected set of inliers 3. go back to 1

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 3/22

Motivation - III
Inaccurate inlier detection for the initial (or subsequent) parameter estimation contributes heavily to the instability of the estimates of the parameters for the remaining models.

0.8 0.6 y 0.4 0.2

0.2

0.4 x

0.6

0.8

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 4/22

Goals
Extend RANSAC to multiRANSAC Present a theoretical analysis of multiRANSAC Show its effectiveness on synthetic and real data sets

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 5/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 6/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 6/22

From RANSAC to multiRANSAC - I

RANSAC key ideas:
Estimate the model parameters using the minimum number of data possible Check which of the remaining data points t the model instantiated with the estimated parameters. Do this a sufciently large number of times (more to come. . . )

RANSAC iteration:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 7/22

From RANSAC to multiRANSAC - II

multiRANSAC key ideas:
At each iteration instantiate models and nd the corresponding CSs Fuse the new CSs with the previously detected ones in a sensible way.

multiRANSAC iteration:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 8/22

From RANSAC to multiRANSAC - II

multiRANSAC key ideas:
At each iteration instantiate models and nd the corresponding CSs Fuse the new CSs with the previously detected ones in a sensible way.

multiRANSAC iteration:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 8/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers?

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 9/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? It can be shown that:

where:

is the number of models is the total number of data points is the number of inliers for the model model is the cardinality of the MSS for the

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 9/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is The probability that this happens for trials is

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is The probability that this happens for Goal: make sure that trials is

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is The probability that this happens for Goal: make sure that Then: trials is

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is The probability that this happens for Goal: make sure that Then: trials is

Problem: a priori we dont know

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

How Many Iterations? - II

The probability that MSSs are not entirely composed by inliers is The probability that this happens for Goal: make sure that Then: trials is

Problem: a priori we dont know

Solution: the estimate for is the cardinality of largest CS found so far.

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 10/22

Fusing the CSs

Observation: the CSs obtained in current iteration can be fused with the CSs obtained in the previous iterations to produce better CSs. More specically: Better is quantied in terms of a tness function
Cardinality MSE

Simplication: just look at the previous iteration Greedy algorithm

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 11/22

Fusing the CSs

Observation: the CSs obtained in current iteration can be fused with the CSs obtained in the previous iterations to produce better CSs.

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 11/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 12/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 12/22

Experiments: Detecting Lines - I

Toy problem: Identify the lines containing the segments.
The stair data set

0.8 0.6 y 0.4 0.2 y

0.8 0.6 0.4 0.2

0.2

0.4 x

0.6

0.8

0.2

0.4 x

0.6

0.8

Wrong estimate

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 13/22

Experiments: Detecting Lines - I

Toy problem: Identify the lines containing the segments.
The stair data set

0.8 0.6 y 0.4 0.2 y

0.8 0.6 0.4 0.2

0.2

0.4 x

0.6

0.8

0.2

0.4 x

0.6

0.8

Correct estimate

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 13/22

Experiments: Detecting Lines - II

Denitions:
Set of correctly detected inliers for the model : the detected set of inliers Percentage of correctly detected inliers:

where

is

models

is the averaged value of over 50 trials and

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 14/22

Experiments: Detecting Lines - II

Denitions:
Set of correctly detected inliers for the model : the detected set of inliers Percentage of correctly detected inliers:

where

is

models

is the averaged value of over 50 trials and Noise std 5.5 6.0 6.5 7.0 7.5 multiRANSAC

sequential RANSAC

95.96% 95.22% 90.08% 90.13% 86.01%

7499 7626 8194 8699 8110

93.80% 86.00% 67.14% 47.92% 37.29%

2016 2049 2080 2103 2103

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 14/22

Experiments: Detecting Homographies

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 15/22

Experiments: Detecting Homographies

multiRANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 15/22

Experiments: Detecting Homographies

sequential RANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 15/22

Experiments: Detecting Homographies

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 16/22

Experiments: Detecting Homographies

, , ,

multiRANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 16/22

Experiments: Detecting Homographies

, , ,

sequential RANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 16/22

Experiments: Detecting Homographies

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 17/22

Experiments: Detecting Homographies

,
,

multiRANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 17/22

Experiments: Detecting Homographies

,
,

sequential RANSAC

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 17/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 18/22

Presentation Overview
Description of multiRANSAC algorithm Experimental Results Conclusions and Future Work

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 18/22

Conclusion & Future Work

We presented the multiRANSAC algorithm On average, better performance than sequential RANSAC Synthetic data (quantitative results) Real data Future Work Quantitative experiments on real imagery Improved CSs fusion strategy Can we detect the number of models? Explore waiting time between updates of the CSs

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 19/22

The End

Thanks for your attention.

Special thanks to Dr. M. Bober, E. Drelie, D. Fedorov and prof. K. Rose for the helpful discussion

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 20/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers?

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? Think in terms of ratio of number of favorable to number of possible

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? The chance of drawing only inliers is given by:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? The chance of having the inliers belonging to the correct model:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? Putting things together we have:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

How Many Iterations? - I

What is the probability that we draw MSSs only composed by inliers? Putting things together we have:

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 21/22

RANSAC vs. multiRANSAC iterations

How big is

in a typical problem?

Number of data Number of models Cardinality of a MSS Probability threshold

Total percentage of inliers Inliers for each class ,

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 22/22

RANSAC vs. multiRANSAC iterations

How big is

in a typical problem?
N=2 5 0
6
W =4 RANSAC multiRANSAC W =3

10

10 T iter

W =4 W =3

W =2

W =2

10

2
W =1 W =1

0.6

0.65

0.7

0.75 r I

0.8

0.85

0.9

The multiRANSAC Algorithm And Its Application to Detect Planar Homographies p. 22/22