Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
Chapter 5 Conclusion and Further Research
5.1 Conclusion
From what we have said above, we can come to the conclusion and summarize
the statements as follows:
1. Concerned experts suggest that the base line is equal to one fortieth or one
fiftieth of the distance between the primary object and camera [12, 14]. From our
experiment we find base line is not a sensitive parameter in stereoscopic photograph,
but the farther the object the more the base line is the rule.
2. In Table 5, we used analysis of variation (ANOVA) to test significance. We
find that F value is less than critical value in 0.05 level of signification, so we accept
the null hypothesis. It has been confirmed that the proper numbers in the scence are
not significantly different.
3. The Eq. 2.3 is in very good agreement with the experiments as shown in
Figures 8 and 9. We can calibrate the camera easily as shown in Section 4.3. In
Section 4.4 we can find that the different digital camera has different focus, in our
experiment for Casio-4z it is 24.028 and for Pentax-S5i it is 25.637. The equations
of regression are zd=25.637×b for Pentax-S5i and zd=24.028×b for Casio-4z. We
32
Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
can use the equations to estimate the distance from camera to object in the scence.
4. In Table 8, we found that the predicted error rate between training and
validation are quite near. The error rate of validation points is between 1.2% and
2.7%.
5. As to relation between error rate and b, it is shown as in Figure 14. We
found that when b at 30 or 71, the error rate is lower.
6. In Table 9, we find that the results of estimate in Figure 16 are reasonable. But
further points are unexpcted, because b value within one to eight for far objects may
cause bigger error. In Table 10, we used b value within five to thirty-five and find that
the predicted error rate lie in between 0.45% and 5.96%.
5.2 Future Research
After the depth estimation, we can domainate its application in our further
research. We would like to mention some of its topics as fellows.
5.2.1 Projection techniques
We may consider projection screen geometries before projection. We need to
know the distance between the screen and projector. Preprocess of the images
before the projection are shown in Figure 18.
5.2.2 Multi-projector displays
Multi-projector displays can be overlapped or not. Corrected projection from
33
Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
each of the multi-projectors are beautiful as shown in Figure 19.
(a) Projection screen geometries (b) Projection in curved screens
(c) Shape adaptive projection (d) Cartoon dioramas Figure 18. Activity of projection techniques
5.2.3 Sparse structure reconstruction
The work of structure reconstruction consists of feature selection, feature
correspondence, camera calibration and Euclidean reconstruction as shown in Figure
20.
34
Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
(a) The examples of multi-projector displays
(b) Type of multi-projector displays Figure 19. The examples of multi-projector displays
(a) Sparse structure reconstruction (b) Investigate of a traffic accident Figure 20. The examples of structure reconstruction
35
