PLEASE NOTE: In order to post on the Board you need to have registered. To register please email paul@sexyloops.com including your real name and username. Registration takes less than 24hrs, unless Paul is fishing deep in the jungle!

Analysis and Observations on Fly Cast Video Footage

Moderator: Torsten

Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Analysis and Observations on Fly Cast Video Footage

#1

Post by Dirk le Roux »

It may be interesting, and of some use towards a more complete understanding of line flight dynamics, to have a thread dedicated to the analysis of fly cast videos and observations on these.

Analysis of different cast types may reveal aspects not recognised before, as well as prevalent differences and similarities. I will start with an analysis and observations on a “normal” horizontal cast’s footage kindly received from Graeme.

Anyone is welcome to add their own analysis to the thread. Workload allowing, I am also available to analyse usable footage anyone may choose to share with me. If you have footage to share, please PM me if you don’t yet have my private email address. A wish list for footage:
  • An even as possible background, contrasting with the colour of the fly line used. Graeme’s method works well, using a dark wall as a background and casting a light-coloured fly line in the sunshine.
  • Markers on the fly line, at same intervals (every 1m works fine), in a colour contrasting with that of the line. Without markers, it is not possible to track points on the line (other than line end and fly) and analysis will be limited.
  • If possible, a leader which is visible against the background will be great.
  • It would be great if some form of regular horizontal intervals could be part of or placed against the background, to aid in correcting the distortion that affordable equipment is likely to produce.
  • Keep full cast extension in the visible frame as far as possible.
  • Focus as sharp as possible.
  • Slow-motion footage (preferably at least 240 fps) at a resolution as high as possible.
  • Footage must please be as “raw” as possible, i.e. not compressed and thereby skipping or repeating frames.
  • Footage should be taken with a sturdily mounted camera in order to be able to track motion at all, as well as to limit camera shake. Also, the camera should preferably be mounted plumb to help distinguish up from down.
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#2

Post by Dirk le Roux »

GRAEME’S NORMAL CAST:

The footage consists of a back-cast and a forward cast. It appears that if there was any wind, it must have been quite light – no grass movement can be seen in the footage. The footage had an area of poor line marker visibility in the right-hand area, especially the lower right-hand quadrant, which made it impossible to locate line markers there and consequently the line marker paths of the forward cast are incomplete. There were enough glimpses of the line in that area to trace line configuration though.

Both the forward cast and the back-cast start with the line in the air and sloping down from the rod tip, quite below the horizontal. It appears the sequence overall has a slant which is lower on the back-cast and higher on the forward cast.

outline_1.jpg
In this image, the red track represents the path of the rod tip, yellow and green tracks that of the fly and the blue and pink tracks that of the loop nose. The white arrow indicates the two RSP positions and the peach coloured arrow is a line parallel to the slope between the two RSP positions.

Both loop nose paths show a “down-up-down” tendency, the forward cast’s being more pronounced and only its upward portion being indeed upward.

The forward cast was more energetic than the back-cast.

A hint of the dolphin nose evident with the back-cast was already present at MCF. Certainly, the back-cast loop initially was narrower than that of the forward cast. This image presents a same-scale comparison of both casts at MCF:
loop sizes BC FC_1.jpg

The back-cast rod tip stop was followed by a short drift and then a slow start of the forward stroke (creep) while the fly leg was still rolling out behind. The forward cast rod tip stop was followed by drift only.
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#3

Post by Dirk le Roux »

INITIAL CONDITIONS FOR THE BACK-CAST:

Line slope early during the stroke:
BC slope.jpg

Line configuration at RSP is a slim S-curve:
BC config.jpg

Rod tip position at RSP is further ahead of the highest tangent-to-the-horizon part of the stroke path than that of the forward cast, as we shall see later, which may provide a clue of launch trajectory.

Absolute velocity (speed) of the rearmost line marker at around RSP for the back-cast was 15.15m/s.



TRACKED PATHS OF MARKERS DURING THE BACK-CAST:

The image below shows the tracked paths of all line markers from the early phase of the casting stroke to past turn-over of the last line marker:
BC markers from pre RSP.jpg

This image shows these paths tracked from RSP only:
BC markers from RSP.jpg

Note, the instance of this last video snapshot shows the dolphin nose is responsible for the dips in tracked paths before final curvature into the loop (here, dolphin nose dip coinciding with the dip in the blue path).

The paths mimic the configuration of line ahead at the time the loop formed, but stretched out and tilted:
paths vs configuration.jpg

• The peak end-of-cast speed of the rearmost line marker on the back-cast was 14.15m/s. Horizontal velocity at the time was 13.14m/s:
M5 peak velocity.jpg
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#4

Post by Dirk le Roux »

RISE AND FALL OF LINE MARKERS DURING THE BACK-CAST

Starting from RSP and ending for each marker individually at the time it had entered the rod leg, the five markers’ averaged Y acceleration was 5.24m/s² downward. However, including “hang time” when in the rod leg, the five markers’ averaged Y acceleration, overall between RSP to near the end of fly leg roll-out (where spotting markers became impossible), was 2.77m/s² downward.

Furthermore, as shown in the below plots, within markers’ motion history are evident periods of quite high downward and upward (Y component) acceleration:
acceleration history BC markers.jpg

All markers’ plots show a drastic downward peak as the top of the loop is rounded, immediately followed by a drastic upward peak as a point is slowed down in joining the rod leg. The nearer-to-rod-tip markers (M 1 being nearest) show a period of hang after their joining the rod leg, with another clear pulse as a wave in the rod leg passed.

The fly’s Y acceleration plot, showing the frequency and magnitude (note x 10²) resulting from its many small curves (flutter?), was included more to have something fill that spot in the image frame.

The last (M 5) line marker’s track, shown with acceleration vectors turned on, illustrates where along its path the peaks occur. At the three circled bends on the left hand, the higher acceleration occurs where the marker passed through bumps in the line configuration. The circled area on extreme right marks the previously mentioned upward acceleration related to the marker coming to “rest” in the rod leg:
M 5 acceleration vectors.jpg

Longer arrows indicate higher magnitude accelerations. Incidentally, the two longest arrows pointing backward in the above image, where the marker rounded the loop nose, register over 200m/s² acceleration in magnitude. That is about 20g! The highest acceleration of this same marker during the cast stroke was just under 6g.

Below is a compilation of markers’ Y position vs. time plots, with a parabola fit to the last marker’s plot, which, though more like a bell curve, is the plot which most closely resembles a parabola shape. Blue arrows there indicate periods of more than 1g upward acceleration and red arrows indicate periods of more than 1g downward acceleration:
M 2 to M 5 y position-time.jpg
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#5

Post by Dirk le Roux »

BACK-CAST FINGERPRINT TRACE

Here is a trace of the line’s configuration during the back-cast flight, at the same time intervals used in the Tracker plots, starting from loop formation:
composite bc.jpg

It looks like a fingerprint of the cast. In it, the progression of bumps as they propagate through the line can be spotted.

These waves may provide a clue as to the fly leg’s change in orientation from generally / to \ and the loop nose path’s less downward travel coinciding with that change in fly leg orientation. A bump (indicated by the white arrow) that started here:
wave 1_1.jpg

have travelled:
wave 2_1.jpg

to here:
wave 3.jpg

and immediately thereafter
wave 3_1.jpg

the bump started moving more upward:
wave 4.jpg

Note the loop’s simultaneous opening and the fly leg now being downwardly oriented.

Once the bump reached the position shown by 3 above, this wave spatially changed direction:
cross-over_2.jpg

The direction change corresponds to the end of the line having rounded the bottom bump (indicated by a above, and by the yellow arrow in the previous snapshots 1 and 2) or alternatively stated, the bottom bump having propagated out of the system. Notice the (white dot) fly position in the above snapshot.

Bumps present in the initial @RSP line configuration, in this cast, propagate along the line toward the fly end. The following observations appear to be constituent to the evolution of these bumps:
  • While further away from the fly end, their curvature progressively narrows (radius decreasing).
  • Nearing their exit at the fly end, their curvature appears again to widen.
  • A bump’s spatial direction of travel is influenced by that of bumps further behind it, or if none there, by the orientation of the trailing leg.
  • Bumps (in the above image, a and c) at times appear to outrun the system slightly, like the often-seen dangle coming through almost as a second loop on some distance casts.
Points on the line move through the evolving course set by the bumps and resulting line orientation, like Bernd’s diagrams show.

Bernd's analysis.jpg
Bernd's analysis.jpg (76.65 KiB) Viewed 5421 times

While Bernd’s blue momentum arrows represent what should be observed in a moving frame of reference, they need some adjustment to earth frame: While line moves through this configured course, the fly leg system also translates in the direction of the cast. The image below shows the marker M 5 path of travel and its interval positions while rounding-the-course-and-translating, this time with velocity (and thus momentum) vectors showing:
composite bc M 5.jpg
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#6

Post by Dirk le Roux »

Back to the rising (b) bump and its relation in this cast to the loop nose path “falling less”. As can be seen in the fingerprint trace with it shown, the loop nose path is more of curve, having tended slightly upward already before the previously discussed bump’s direction change. What to ascribe this earlier change to?

nose path curve_early_1.jpg

One candidate factor: Soon after the initially quite narrow loop’s formation (1), the loop nose peak shifted more to the top of the loop (2). Another candidate factor: Bump c, its travel direction indicated by the blue arrow at 2 and 3 and influenced by the travel direction of bump b behind it, diverges from the loop nose path.

Note that the bumps do not propagate through the loop. Thereby they don’t influence the loop’s altitude by the effect their passing directly through the loop may have had.

The loop nose path’s more downwards curve toward the end of the cast corresponds with bump b having left the system and the path now being influenced by the downward orientation of remaining fly leg:

nose path curve_late_1.JPG

Analysis and observations on the forward cast to follow in a day or so.

Regards,
Dirk
User avatar
gordonjudd
Posts: 1858
Joined: Sat Jan 19, 2013 11:36 pm
Answers: 0
Location: Southern California

Re: Analysis and Observations on Fly Cast Video Footage

#7

Post by gordonjudd »

Footage should be taken with a sturdily mounted camera in order to be able to track motion at all, as well as to limit camera shake. Also, the camera should preferably be mounted plumb to help distinguish up from down.
Dirk,
Thanks for taking the time to digitize the hundreds if not thousands of points involved with getting marker positions on high speed video files. People who have not been through that tiresome process have no idea of the dedication and time it takes to produce your plots.

In terms of camera positioning I would add that the camera should be pointing 90 degrees to the path of the cast. That will help to minimize the distortion of the apparent x position data, although some non-itineraries will always result especially when you use shorter focal lengths in the camera to see the entire fly path. Correcting for those camera distortions is a complicated process that still remains a mystery to me.

With that perpendicular alignment you will generally only be able to see the casting arm shoulder. Being able to see Graeme's back in those videos indicates the forward cast was going away from the camera, and thus foreshortens the apparent x marker positions. That may be why the loop velocity values you are getting (around 7 m/s) are so low.

Taking derivatives go get velocity and acceleration values are difficult to do when a simple delta_d/delta_t computation is made with just using adjacent points. When the delta_t value is small as it is with 240 fps videos then small digitizing errors can really produce a lot of ripple in the resulting velocity values. There are spline fitting routines available to smooth out those derivatives that you might try to find so that you get some meaningful velocity and acceleration data. When I see ripple in derivative data I attribute it to measurement errors not something that is really going on at such small time scales.

This is the kind of data we need to start getting a better idea of loop propagation and I congratulate you for investing all of the time you have spent in producing this meaningful analysis.

Gordy
Torsten
Posts: 455
Joined: Tue Jan 22, 2013 7:34 pm
Answers: 0

Re: Analysis and Observations on Fly Cast Video Footage

#8

Post by Torsten »

Nice topic Dirk,

I think it's very useful to have guidelines for recording and digitizing fly casting videos. So perhaps we can collect ideas, how-to instructions and write something together, e.g. like a picture story. Relevant aspects are in my opinion:

* Video Equipment List, what do we need to start recording, are some video cams recommended? What about action cams?

* Setup (like post #1)

* What is the best software to digitize these videos? ( above looks like https://physlets.org/tracker/ (?) )

* Steps to digitize the fly line, is this a full manual process?

* Sharing of digitized data

etc..

--

I'd like to see some reference casts digitized - like the typical wedge shaped loop in the Sexyloops logo, static rolls and switch casts.
This can be used then to improve models and the understanding of fly casting in general.

Thanks,
Torsten
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#9

Post by Dirk le Roux »

Thanks, Gordy
Correcting for those camera distortions is a complicated process that still remains a mystery to me.
That is why I proposed some regular distance interval indication in the setting. A visible tape on the ground, along the length of the cast, might do. Working with footage without that, we need to keep in mind that an unknown amount of distortion is likely to be present. Displacement-dependent values, especially if the action takes place over the full view frame width, therefore should be taken with a pinch (of unknown size) of salt. The trends, however, can still be read from the available data results. I think the Y displacement values in the cast analysed should not have suffered too much from camera distortion, due to the fairly narrow vertical band within which movement took place.
Being able to see Graeme's back in those videos indicates the forward cast was going away from the camera, and thus foreshortens the apparent x marker positions. That may be why the loop velocity values you are getting (around 7 m/s) are so low.
Graeme simply stood less square to the cast, facing the wall and casting across his chest. I am fairly convinced the cast was shot perpendicularly to Graeme's position relative to the wall (and the cast). Graeme can confirm. Regarding the velocity values, remember that this was a very laid-back kind of cast.
Taking derivatives [t]o get velocity and acceleration values are difficult to do when a simple delta_d/delta_t computation is made with just using adjacent points. When the delta_t value is small as it is with 240 fps videos then small digitizing errors can really produce a lot of ripple in the resulting velocity values. There are spline fitting routines available to smooth out those derivatives that you might try to find so that you get some meaningful velocity and acceleration data. When I see ripple in derivative data I attribute it to measurement errors not something that is really going on at such small time scales.
In my experience with Tracker, smaller time steps lead indeed to noisy derivatives and larger steps are sufficient to get a picture of what goes on without being too inaccurate and without losing perhaps important detail. With this analysis I plotted points at 6-frame steps, so the delta_t was 1/40th s as opposed to 1/240th. I think that gave serviceable results, not to mention cutting workload by a third!

Here is an Excel-consolidated chart showing both 2-frame-step and 6-frame-step data for M 5 marker's absolute velocity gained from Tracker:

M 5 excel overlay.jpg

You can see the 6-frame-step plot (red) has less jitter than the 2-frame-step one (blue) and corresponds quite well with a 7-period moving average trendline (green) of the latter.

Or I can read that plot like this:
M 5 excel overlay_1.jpg

Horizontal speed small variations were not part of my focus for this analysis in any case. I merely included the two M 5 plots as an additional perspective. I do think though that the vertical acceleration plots represent reality fairly closely, as their ups and downs can be correlated by the marker path curves.

Regards,
Dirk
Dirk le Roux
Posts: 556
Joined: Thu Sep 19, 2013 5:09 pm
Answers: 0
Location: Pretoria, South Africa

Re: Analysis and Observations on Fly Cast Video Footage

#10

Post by Dirk le Roux »

Hi Torsten

* Video Equipment List, what do we need to start recording, are some video cams recommended? What about action cams?
Others can comment, but I think within people's budgets smartphone produced high-resolution 240fps footage is OK. A good stand/tripod may be of more importance.

* Setup (like post #1)
Add some yardsticks to the scene to enable quantifying distortion! If the topic is not a full-length cast, shoot perpendicular to the area of interest, like Graeme's been doing with his "death by tell tail" work.

* What is the best software to digitize these videos? ( above looks like https://physlets.org/tracker/ (?) )
Yes, Tracker is what Graeme and I (from his introduction to this wonderful free tool) use.

* Steps to digitize the fly line, is this a full manual process?
Auto tracking with Tracker is possible but I haven't yet come across footage which lends itself to success with it. Once you have plotted the points, Tracker can provide all sorts of data.

* Sharing of digitized data
If anyone wants my files, PM me.

Regards,
Dirk
Post Reply

Return to “Flycasting Physics”