0% found this document useful (0 votes)
13 views10 pages

Video Analysis Lab Report on Calibration Errors

The lab report investigates the impact of miscalibration on video analysis of a buggy's motion, revealing a significant error when the calibration object is not properly positioned. The study found a percentage error of 9.4 ± 4.3% in buggy velocity due to miscalibration, compared to the predicted error of 10.6 ± 0.2%. The results suggest that video analysis can be a reliable method for quality control in motion analysis, provided proper calibration is maintained.

Uploaded by

andyduong555
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
13 views10 pages

Video Analysis Lab Report on Calibration Errors

The lab report investigates the impact of miscalibration on video analysis of a buggy's motion, revealing a significant error when the calibration object is not properly positioned. The study found a percentage error of 9.4 ± 4.3% in buggy velocity due to miscalibration, compared to the predicted error of 10.6 ± 0.2%. The results suggest that video analysis can be a reliable method for quality control in motion analysis, provided proper calibration is maintained.

Uploaded by

andyduong555
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Appendix​ - Sample Lab Report

GOOD SAMPLE LAB


Lab I, Problem 1b: Measurement and Uncertainty Using Video Analysis
Pat Kline
August 7th, 2017
Physics 1301W, Professor: Y. Ryoma, TA: P. Hearn

Abstract

The use of video analysis software to model the motion of an object


and the importance of properly calibrating the video were studied.
A video of a constant velocity electric buggy was made and the
motion was analyzed. With a camera placed 94.0 cm from the
plane of motion, it was found that a calibration object 10.0 cm in
front of the plane of motion led to an error of 9.4 ± 4.3 % in ​the
measured buggy velocity compared to having the calibration object
in the plane of motion​. ​Additionally, it was found that the random
uncertainty of the measured velocity using video analysis was less
than that from a previous hand measurement made using a dowel
and stopwatch.

Introduction
BuggyMagic wants to use video analysis as a method of quality control to provide
buggies of constant velocity. The quality-control team have been analyzing videos of
buggy motion and have gotten different velocity values, even for the same video. The
team suspects that reaching a different conclusion using the same video may be due to
miscalibration. Therefore, to demonstrate the effects of miscalibration, a buggy’s motion
will be analyzed when a calibration object is in the plane of motion compared to when
the calibration object lies in front of the plane.

Prediction
Using simple geometry, the effects of miscalibration can be estimated. As seen in Figure
1, the whole setup can be modelled by a triangle ABC with A representing the
point-like camera and the side BC representing the plane of motion, or a part of it. If a
calibration object, signified by the line DE, is added in front of the plane of motion, the
camera thinks that the plane of motion is shorter than it actually is.
APPENDIX: SAMPLE LAB REPORT

Figure 1: Sketch of a setup with miscalibration. The corner A is the camera, the side BC
is the plane of motion and the side DE is the calibration object. This setup makes the
camera think that the length |BC| is equal to the length |DE| .

The percentage error in distances due to the miscalibration is then predicted to be


|BC|−|DE| |DE|
%E = |BC| =1 − |BC| (1)

Since the length |BC| does not correspond to any real object and the experiment could
also be done with a different calibration object of random length, a different way of
presenting the prediction is preferred. Using simple geometry of similar triangles, the
distances between the plane of motion, the camera, and the calibration object can be
related through
|DE| |AD|
|BC|
= |AB|
(2)

and, along with the relation |AD| + |DB| = |AB| , Eq.1 can be rewritten as
|DB|
%E = |AB| (3)

Thus, the effect of miscalibration can be predicted based on the distances of the camera
and the calibration object from the plane of motion.

Procedure
Two videos of a buggy’s motion were made. The camera was located 94.0 cm away
from the plane of motion. A wooden dowel of length 22.0 cm was used as a calibration
object for both videos: one in which it was aligned with the plane of motion and another
where it was placed 10.0 cm in front of and parallel to the plane of motion. The buggy
was released to move with a constant velocity in the plane of motion for a distance of
about 5 buggy lengths to accumulate enough data points. MotionLab analysis software
was used to generate (horizontal position, time) pairs at each frame in the trajectory
and, by linear interpolation, (horizontal velocity, time) pairs between each pair of
consecutive frames in the trajectory. To analyze the motion in detail, data was exported
APPENDIX: SAMPLE LAB REPORT

into Excel and plots were generated of the buggy’s position and velocity as functions of
time.

Data
The distance from the camera to the plane of motion was
|AB| = 94.0 ± 0.2 cm
The distance the dowel was placed in front of the plane of motion was
|DB| = 10.0 ± 0.2 cm
Then the expected percentage error in the buggy velocity from the miscalibration is
%E = 10.6 ± 0.2 %

As seen in Figures 2 and 3, the data was exported into Excel to produce graphs of the
data.

Figure 2. The measured position of the buggy as a function of time. Equations of best fit
lines through the data show the velocity as the slope value. ​Error bars are not shown
because the uncertainty of individual point values was small compared to the
systematic error.
APPENDIX: SAMPLE LAB REPORT

Figure 3. The measured velocity of the buggy as a function of time for the correctly
calibrated (triangles) and miscalibrated (circles) videos, along with mean velocities and
uncertainty ranges: v c = 38.96±0.87 cm/s , v m = 35.31±0.82 cm/s . Error bars are not shown
because the uncertainty of individual point values was small compared to the
systematic error.

Analysis
The results of the buggy’s motion are shown in Figure 2 (position versus time) and
Figure 3 (velocity versus time). The first thing to note is the early velocity values for the
proper calibration in Figure 3. This large deviation from the mean is due to taking data
points as soon as the buggy was started. Pressure from the hand holding the buggy
slowed it down until it was fully released. This error can also be seen in Figure 2 as a
slight curve in the beginning. All analysis of the data excluded those early points. This
is not seen on the miscalibrated data as care was taken to start data acquisition after the
buggy was fully released.

Based on our prediction we expect the velocity percentage error from the miscalibrated
data to be 10.6 ± 0.2 % . From Figure 3, we found the velocity and the average deviation
for the proper calibration case to be v c = 38.96±0.87 cm/s, while the velocity for the
miscalibration case turned out to be v m = 35.31±0.82 cm/s . As expected, in a case with
the calibration object placed in front of the plane of motion, the buggy appears to move
slower. From those values the percentage error of those velocities is calculated to be
%E = 9.4±4.3% . This is slightly less than expected but comparable when the
uncertainty range from the propagation of error is taken into account.
APPENDIX: SAMPLE LAB REPORT

From earlier hand measurements in Problem 1a, the buggy velocity was found to be
v h = 40.2 ±3.2 cm/s; therefore, the result achieved by the properly calibrated video
does fall within the expected uncertainty of v h .

In comparison with the hand measurements, the random measurement uncertainty is


smaller for the video analysis technique. Considering the uncertainty of a few pixels in
distance measurements and some small fraction of a second in time measurements, it
can be said that the random uncertainty in the velocity value found in MotionLab is
negligible, particularly compared to the larger error values that a possible
miscalibration procedure might introduce. Thus, video analysis is a more effective
technique for investigating motion, provided that the proper calibration is used.

Several potential sources of error might explain the difference between the predicted
and the measured percentage error values. One is the distortion effect of the camera;
data points that do not come from the center-most portion of the field of view might be
discarded to limit this effect. Another is the slightly curved trajectory of the buggy. This
effect was minimized by modifying the initial direction of the motion. Moreover, to
prevent the buggy’s wheels from hitting the dowel, the calibration object could not be
placed exactly on the plane of motion, which might have caused some minor error.

Conclusion
It was shown that a significant error can appear while using video-analysis software if
the video is not properly calibrated. The measured velocity of a buggy when the video
was properly calibrated with the calibration object located in the plane of motion was
compared to the measured velocity when it was miscalibrated with the calibration
object in front of the plane of motion.

The effect of miscalibration was modelled using simple geometry of triangles. Even this
simple model was able to predict the percentage error in the velocity caused by
miscalibration, 10.6 ± 0.2 % , within the uncertainty value of the measured percentage
error, 9.4±4.3% .

This result implies that it is feasible for BuggyMagic to utilize a video analysis
technique for quality control, provided that the quality control team properly calibrates
their videos of the buggies’ motion.
APPENDIX: SAMPLE LAB REPORT

BAD SAMPLE LAB

Lab I, Problem 1b

Comte de Rochefort

August 7, 2017

Introduction
We seek to determine how miscalibration effects the results of a video analysis in
MotionLab. To do this, we used a buggy that moves with a constant velocity. We used
iPad to record videos of it moving on a straight line, being as careful as possible to
simulate the constant velocity motion accurately and to minimize errors. We analyzed
the videos with MotionLab, taking several data points for each case: with a proper
calibration and a miscalibration.

Prediction

|DB|
%E = |AB|
Procedure
We performed this experiment by a scientific procedure. We first made a prediction;
then, we performed the experiment; then, we analyzed the data; then, we drew a
conclusion.

We began by gathering the materials. They included:


• meter stick
• dowel
• buggy that can move with a constant velocity
• iPad on tripod
• computer
• tape

We put the dowel right next to the wheels for the proper calibration and 10 cm in front
of the wheels for the miscalibration of MotionLab. We faced the camera toward the
buggy.
APPENDIX: SAMPLE LAB REPORT

We released the buggy after turning it on and recorded a video using the iPad. We then
analyzed the motion using MotionLab. This began with the proper calibration. We first
set time zero at the exact time when we released the buggy. We then had to calibrate the
length. We put the dowel in the frame of the video, so we used the dowel for the
calibration. We then defined our coordinate system so that the motion of the buggy
would be straight to the left.

We then made predictions about the motion. We predicted that the y position would
not change and that the x position would be a straight line with a slope B=38cm/s. The
predicted equations were y(z)=0 and x(z)=-38z.

We then had to acquire data. We measured the position of the buggy at each frame in
the video, starting at t=0. We put the red point at the center of the front wheel each time
for consistency. This was important to keep from measuring a length that changed from
frame to frame based on where we put the data point on the buggy. We also did not use
some of the frames at the end of the video, where the buggy was at the edge where the
camera is susceptible to the fisheye effect.

When this was finished, we fit functions to the data points. The functions did not fit the
points exactly, but they were acceptably close. We fit y(z)=0 for the y position and
x(z)=-37.9z for the x position. These were close to our predictions.

It then came time to make predictions of the velocity graphs. We predicted that the Vy
graph would be a straight line with Vy(z)=0 and that the Vx graph would be a linear
line with Vx(z)=-37.9.

Next, we fit the functions to the data points for the velocity graphs. We got the
predictions exactly right.

We then printed our data for the constant velocity buggy and closed MotionLab.

We repeated this process for the miscalibration case. It was mostly the same, with some
exceptions. The x(z) fit was x(z)=-35z instead of y(z)=-37.9z. The Vx(z) prediction was
Vx(z)=-35 instead of Vx(z)=-37.9. These were also exactly right, so the Vx(z) fit was the
same.
APPENDIX: SAMPLE LAB REPORT

Data
APPENDIX: SAMPLE LAB REPORT

Analysis
We can calculate the velocity from the MotionLab fit functions. To do this, we use the
formula x = x0+v0t+1/2atˆ2. Then v is just the coefficient of z in the position fits. This
gives us
the proper calibration: v=37.9
the miscalibration: v=35
APPENDIX: SAMPLE LAB REPORT

The velocity can also be calculated using the formula v=v0+at. Then v is just the
constant in the velocity fits. This gives us
the proper calibration: v=38.8
the miscalibration: v=35.2

We know that the velocity found in the video analysis with the proper calibration is 37.9
cm/s, so we need to compare the velocity value we get with a miscalibration procedure
to this number. Looking at the data from the fits, we can see that they are close to each
other, so the error in this lab must not be significant.

We need to analyze the sources of error in the lab to interpret our result. One is human
error, which can never be totally eliminated. Another error is the error in MotionLab.
This is obvious because the data points don’t lie right on the fit, but are spread out
around it. Another error is that the meter stick could only measure the distances to
+/-0.05cm. There was error in the fisheye effect of the camera lens. The buggy didn’t
move on a straight line exactly but we tried to fix it.

Conclusion
We predicted that %E would be 10.6%, and what we measured, 7.7%, is very close to
this. The errors are therefore not significant to our result. We can say that the error due
to miscalibration can be calculated as in prediction section. This experiment was
definitely a success.

Common questions

Powered by AI

The effects of miscalibration in video analysis are modeled using simple geometry, specifically similar triangles. The setup with the camera and the object is represented as a triangle, where the effect of miscalibration can be estimated through the relationships between the sides and angles of this triangle . This model predicts errors based on the relative distances of the camera and calibration object from the plane of motion .

Miscalibration causes video analysis software to perceive the motion plane as shorter than its actual length, resulting in underestimated velocity values. For instance, the velocity was measured lower at 5.31±0.82 cm/s when miscalibrated compared to 8.96±0.87 cm/s with proper calibration . This implies that for accurate quality control, strict adherence to calibration protocols is critical to ensure consistency and reliability in velocity measurements .

Apart from miscalibration, potential sources of error in video analysis include human errors during measurement, the fisheye effect from the camera lens, and the non-linear movement of the object not being strictly straight . Additionally, systematic errors might occur if the calibration object cannot be perfectly aligned with the plane of motion or if the camera distorts distances outside its central view .

Placing a calibration object in front of the plane of motion rather than within the plane can lead to significant errors in the perceived velocity of an object. For example, when a calibration object was placed 10 cm in front of the plane, the measured velocity showed an error of 9.4±4.3%, as opposed to when it was correctly calibrated in the plane, where the velocity was more accurate . The source of this error is the camera perceiving the plane of motion as a shorter distance than its actual length, thus affecting the velocity calculation .

The predictions based on triangular geometry closely match the observed data within uncertainty ranges. The percentage error, predicted to be 0.6 ± 0.2 %, aligns with measured values, although actual measured error (9.4±4.3%) may slightly deviate due to factors like camera distortion or positioning imprecision. Despite these discrepancies, the geometrical approach provides a reliable estimate for potential error when appropriately applied .

Video analysis techniques improve measurement accuracy by greatly reducing random uncertainties compared to traditional manual methods. For instance, video analysis minimizes the randomness due to human error in timing and capturing precise distances, offering more precise measurements within carefully calibrated environments . The method also reduces the influence of minor time and distance inaccuracies, making it superior for consistent velocity studies .

To minimize initial velocity deviations, data acquisition was started after the buggy was fully released, eliminating the initial pressure from releasing the buggy that was slowing it down. This approach helps to ensure a more accurate capture of the buggy's constant velocity .

The starting direction impacts the accuracy of trajectory data due to potential curvatures in motion; this was addressed by adjusting the initial direction to ensure a straight-line path. Such adjustments are critical to prevent trajectory distortions from affecting velocity measurements, confirming that straight alignment can mitigate these errors .

The placement of the calibration object dramatically affects the error percentage in measured velocity. When the dowel is placed in front of the plane, the error is calculated to be 9.4±4.3%, compared to a correctly placed calibration yielding more accurate values. This demonstrates how crucial correct spatial alignment is in minimizing error percentage .

Video analysis tends to have smaller random uncertainties compared to manual measurement techniques such as using a dowel and stopwatch. The random uncertainty in video analysis is almost negligible when compared to the larger systematic error introduced by potential miscalibrations . This makes video analysis a more precise method for motion studies, assuming proper calibration is used .

You might also like