Metals: Steel prices ski jumping off steepened Chinese slope - Motilal Oswal
High-speed video image analysis of ski jumping flight posture
Transcript of High-speed video image analysis of ski jumping flight posture
ORIGINAL ARTICLE
High-speed video image analysis of ski jumping flight posture
Masahide Murakami • Masato Iwase •
Kazuya Seo • Yuji Ohgi • Reno Koyanagi
� International Sports Engineering Association 2014
Abstract Ski jumping flight posture was analyzed for
achieving large flight distance on the basis of high-speed
video images of the initial 40 m part of 120-m ski jumping
flight. The time variations of the forward leaning angle and
the ski angle of attack were measured from the video
images, and the aerodynamic forces were calculated from
the kinematic data derived from the images. Some corre-
lations were investigated between the initial-speed cor-
rected flight distance and such parameters as the angles of
jumper, the initial transition time and the aerodynamic
force coefficients. The result indicated that small body
angle of attack was a key for large flight distance in the
initial phase of flight because of small drag force, and that
the most distinctive fault of beginners was too large body
angle of attack and ski angle of attack leading to aerody-
namic stall. Too small drag force does not give an optimal
condition for large flight distance because the lift force is
also too small. The ratio of the lift to the drag was larger
than 0.95 for advanced jumpers.
Keywords Ski jumping � Image analysis � Flight
distance � Jumper posture � Aerodynamic force coefficients
1 Introduction
Ski jumping, in particular the flight phase, can be quanti-
tatively treated from the aerodynamics point of view. Some
wind tunnel experiments [1–5], field experiments with the
aid of high-speed image analysis [4, 6] and a simulation
study based on a multisegment dynamic model [7] were
conducted to search for the condition of maximum flight
distance and safety. The wind tunnel test for a V-style
flight was conducted, and a database for the aerodynamic
force coefficients was constructed by Seo et al. [5]. This
was further used in a jumping flight optimization study [8].
The aerodynamic data measured in wind tunnel experi-
ments are the most objective and reliable. But there is not
necessarily accumulation of wind tunnel data enough to
respond to so wide range of variations of parameters and
the data are not available for all parameters. In particular,
they are lacking the cases where the ski-body angle is large
and is rapidly changing in a short time typically found in
the early flight phase from a takeoff to a quasi-steady flight.
Wind tunnel data are not always updated corresponding to
occasional changes in ski length and jump suit due to the
change of rules, and in jumping flight technique from a
parallel flight to a V-style flight. In the image analysis
studies, several angles to characterize the posture of jumper
and the angle of attack were derived, based on which the
optimum jumping posture was inductively discussed. In
relation to this kind of researches, the aerodynamic data
must be referred to in order to give a rational background to
the discussion. However, the optimization could be per-
formed only within the limit of available aerodynamic data
M. Murakami (&)
University of Tsukuba, Tsukuba 305-8573, Japan
e-mail: [email protected]
M. Iwase
Graduate School of Systems and Information Engineering,
University of Tsukuba, Tsukuba 305-8573, Japan
K. Seo
Faculty of Education, Art and Science, Yamagata University,
Yamagata 990-8560, Japan
Y. Ohgi � R. Koyanagi
Graduate School of Media and Governance, Keio University,
Fujisawa 252-0882, Japan
Sports Eng
DOI 10.1007/s12283-014-0157-z
obtained in wind tunnel tests. For the case of simulation
studies, the situation is almost similar to this with respect to
the necessity of wind tunnel aerodynamic data.
In view of these researches, we attempted to derive the
aerodynamic forces exerted on a jumper on the basis of the
image analysis of high-speed video image of real jumping
flights [9, 10] to be complementary with the existing
researches. We expected that the aerodynamic force data
should have been obtained for the parameters under real
flight condition as the result of this study. We furthermore
considered that aerodynamic data even in the cases where
the parameters were rapidly changing on a large scale,
including aerodynamically unsteady situations, could be
acquired in the study. As a result, coaches on site could
more rigorously give the jumpers advice while the video
images of their flights and the corresponding aerodynamic
data are simultaneously shown to them. More scientific and
reasonable discussion could also be developed on the
correlations between the flight distance and such parame-
ters as angles characterizing the flight posture by showing
the aerodynamic data. It was one of the objectives of the
present study that the image analysis procedure was further
improved for higher accuracy by executing detailed on-site
survey of the jumping field, Hakuba Jumping Stadium, and
by applying a more accurate image correction based on the
survey data than our previous analyses [9, 10]. Further-
more, large amount of image analysis data of jumping
flight have been accumulated so that a statistical data
analysis can be made. As the second objective of this study
some correlations were investigated for the optimal flight
posture leading to large flight distance between the flight
distance and such key parameters as the angles of jumper,
the initial transition time to a quasi-steady flight and the
aerodynamic force coefficients. Furthermore, the jumping
performances are compared between advanced and begin-
ner jumpers to extract typical faults of beginner jumpers.
2 Methods
The flights of large-hill class jumping were recorded with a
high-speed video camera (the main camera) (Photron
FASTCAM 1024PCI) set on the top of the coach tower of
the normal-hill jumping field that is built almost in parallel
to the large-hill jumping field 50 m away. The location of
the camera is about 15 m downward from the takeoff point,
aside the deck of the normal-hill landing slope, and 65 m
away from the center line of the large-hill landing slope.
The field of view of the camera with a wide-angle lens with
a focal length of 28 mm covers the upper 40 m range of the
large-hill landing slope, which roughly corresponds to an
initial flight portion for 2 s. The video camera images were
taken at the rate of 250 frames/s. An auxiliary video
camera (Photron FASTCAM 1024PCI) was also used for
the purpose of calibration of the angle data of a jumper.
This camera was set on the top of the large-hill coach tower
20 m aside the deck of the large-hill landing slope. The
jumper image taken with this camera is more than six times
larger than that of the main camera so that the forward
leaning angle can be measured with an accuracy sufficient
enough to calibrate the angle data of the images taken with
the main camera. The subjects in the field tests consisted of
six jumpers of Japanese Championship level including a
number of Japanese Olympic Team jumpers, 10 jumpers of
university and high school students and a number of World
Cup higher rank jumpers. Total numbers of flights that
were recorded are about 100. However, for some flights,
the data of the takeoff speed and/or the flight distance were
lacking, and thus these data were not analyzed. As no wind
data were available during the field tests, the effect of wind
was ignored. As a matter of fact, wind was weak to mod-
erate during the tests in the mid summers of 2008 and
2009. The coordinate system, and the centre of gravity
(CG) and the angles characterizing the posture of a jumper
are shown in Fig. 1; a is the ski angle of attack, b the flight
angle, c the body angle of attack, h the jumper forward
leaning angle, U the flight velocity. The angle h is defined
as the angle between a ski and legs. The bend angle of
waist could not be measured owing to smallness of a
jumper in the high-speed video images. Among these
angles, h is readily measured from a single instantaneous
frame of the video record, while the other angles must be
calculated after an image analysis for several sequential
images of a jumper. It is because they are the angles that
are defined with respect to the flight velocity vector U of
the centre of mass of a jumper. The flight angle b is the
angle of U measured from the horizontal line, and a and care angles measured from the vector U. In order to derive
the flight vector, and consequently the flight speed and the
acceleration, the time sequence of the CG of a jumper
under flight is pursued in the video images. In this study,
Fig. 1 Coordinate system and the definition of angles characterizing
the posture of a jumper
M. Murakami et al.
the CG value was calculated as a kind of weighted average
of the positions of jumper waist, head and toe. The specific
weights of body segments of a jumper were calculated by
referring to the paper [11]. The image tracking for these
selected positions of a jumper was mostly accomplished
automatically with the aid of image analysis software
(TEMA lite), but partly with some manual assist for cases
where the target positions were hard to be distinguished
because they seemed as if they were blended into the
background. The mass of a jumper was assumed to be
70 kg including ski. The coordinate origin is the takeoff
point, and the x- and y-axes are taken horizontally and
vertically downward, respectively. The z-axis is taken to
the horizontal depth direction. We used a left-handed
coordinate system. The aerodynamic forces, the lift L that
is perpendicular to U and the drag D that is in the opposite
direction to U, are computed in terms of the accelerations
in the x and y directions, ax and ay, and the flight angle bafter the effect of gravity g is subtracted.
L ¼ m ax sin b� ay � g� �
cos b� �
ð1Þ
D ¼ �m ax cos bþ ay � g� �
sin b� �
; ð2Þ
where m is the jumper mass. The lift area SL and the drag
area SD are conventionally used instead of the lift coeffi-
cient and the drag coefficient because the projected area of
a jumper largely changes during a flight [1–3].
SL ¼ 2L=ðqU2Þ ð3Þ
SD ¼ 2D=ðqU2Þ; ð4Þ
where q is the air density, and U is the air speed or the
jumper speed.
A single fixed video camera with a wide-angle lens was
used instead of adopting a three-dimensional camera system.
Some image distortion that is unavoidable in our simple
camera system must be corrected before quantitative image
analysis. For this purpose, site survey had been executed for
24 points on the centre line of the large-hill landing slope by
using a laser-surveying instrument. For the purpose of cali-
bration of position information we took the images of staffs
with four markers stood right on the 24 points surveyed on the
centre line of the landing slope with the video camera situated
at the same location as the jump image recording. The four
markers were mounted in 1 m intervals in the vertical
direction on the staffs. Thus, the positions of total 96 spatial
points were identified in the air above the landing slope on the
video image, forming a pixel map. Based on these points in
the images, of which coordinate values are known, we con-
structed an piecewise interpolation formula system for
conversion from the pixel coordinate (X, Y) to the physical
x–y coordinate. They are polynomials of 2nd order of X and Y
for the y-coordinate, and of 4th order for the x-coordinate.
This calibration procedure was done every day prior to the
video shoot in order to deal with minute changes of the
camera position and camera angle. The maximum values of
overall errors in the positions in an image are estimated to be
37 cm in the x-direction and 5 cm in the y-direction. The
breakdown of these values is as follows: a spatial resolution is
40 mm (=40 m/1,024 pixels); the maximum deviation of a
jumper in flight from the center line of the landing slope is
estimated to be 1 m in the z-direction, which is an interpo-
lation from the deviation at the landing, within ;2 m. This
1 m deviation may cause an error in the x-position 30 cm; the
positioning accuracy in the interpolation procedure is about
2.4 cm in the x-direction and 0.8 cm in the y-direction,
respectively. For the data analysis, the first and the second
order time derivatives of the CG data were computed for the
velocity and acceleration, from which the aerodynamic forces
are derived. The first order numerical differentiation scheme
of the 3rd accuracy for the sequential data of the CG coor-
dinate fi is adopted for the calculation of the velocity.
df
dt¼ �fþ2 þ 8fþ1 � 8f�1 þ f�2
12Dtð5Þ
Here, f? 1 is the value f at the instant of i ? 1 and Dt is the
time interval of the sequential data. It is, however, obvious that
the scattering of the time-series data of the CG is very large.
Thus, for the calculation of the velocity the time-series of the
first order time derivative data were, first, low-pass filtered
with a cut-off frequency of 6 Hz, and then the weighted least
square regression with a higher order polynomial are piece-
wise applied for smoothing. For the calculation of the accel-
eration, the smoothed velocity data were numerically
differentiated, low-pass filtered and finally smoothed.
3 Results and discussion
3.1 Derivations of the aerodynamic force coefficients
and the angles characterizing a jumper posture
Typical examples of the time variations of SL and the ski
angle of attack a are shown in Fig. 2a for two ski jumpers,
a Japanese A-class jumper (flight distance of 131 m) as an
advanced jumper and a high school student jumper (102 m)
as a beginner jumper. The takeoff speed was almost same
for the two cases. It is seen that SL is larger for the
advanced-class jumper than that for the high school student
jumper throughout the record. It is interesting to note that
SL begins to decrease a little before 1 s for the advanced-
class jumper and around 0.7 s for the high school student
jumper. This must be a result of aerodynamic stall. The
stall started earlier for the high school student jumper
because of larger a as a result of excessive pitching motion
of his ski (a). As seen in the previous wind tunnel mea-
surement result [5], the stall angle is between 35� and 40�,
Image analysis of ski jumping flight posture
though it is hard to exactly define the stall angle owing to
the three-dimensional nature of a jumper body. The
symptom of stall, which is the decrease in the rate of
increase of the lift and the increase in the rate of increase of
the drag, begins to appear around 25 degree of the ski angle
of attack a for the high school student jumper. The results
of the body angle of attack c and the body forward leaning
angle h are shown in Fig. 2b and c respectively. For these
figures, the data were shown also for both the advanced-
class and the high school student jumpers. It is seen in
Fig. 2b that c is larger for the high school jumper to result
in larger drag due to his posture where it nearly stands up
against airflow. Figure 2c shows that h values are not so
different between the two jumps, though the flight dis-
tances are greatly different. However, the time tf at which a
flight entered a quasi-steady flight phase, which is the
phase of metastable flight without change of a large-scale
posture or angles, as can be simply defined in Fig. 2c, is
different for the two flights, 0.4 s for the advanced jumper
and 0.8 s for the beginner jumper. The jumping of
advanced jumper enters the quasi-steady flight phase ear-
lier than the high school jumper. In the early flight phase
before tf, the body nearly stands up against airflow resulting
in larger c causing larger drag. Therefore a quick transition
to a quasi-steady flight phase with small c is desirable for a
successful jumping flight. In a beginner’s jump the angle ccannot be made small even in the quasi-steady phase.
3.2 Correlations between the flight distance
and the jumper posture parameters
3.2.1 Calculation of the normalized flight distance
for initial speed compensation
It is understood that the aerodynamic force coefficients
vary as functions of jumper posture characterized by the
angles, a, c and h, which, in turn, govern the flight distance.
We investigate the correlations among them. One of the
most important measures for the evaluation of jumping
performance is the flight distance Dj, which is, however,
directly influenced by the takeoff speed U0. The magnitude
of U0 depends on the start point and the crouching posture
in the in-run and waxing, which were individually out of
control in the present field tests. Therefore, the flight dis-
tance must be compensated for the effect of U0 prior to the
detailed data analysis. For this purpose, the reference flight
distance, Dv, is calculated as a function of U0, which is the
flight distance without any effects of aerodynamic forces,
which is equivalent to the free flight distance of a point
mass in vacuum. The normalized flight distance, Dr, is
defined by Dr = Dj/Dv, which is considered to be almost
free from the effect of the takeoff speed U0. The specifi-
cation of the virtual jumping field for the calculation of Dv
is: 11 degree of the downward angle at the takeoff point,
the height of the takeoff point = 4 m, the landing slope
with an angle of 37.5� immediately follows the takeoff
point without a deck. It should be noted that the present Dv
value is slightly larger than the possible free flight distance
in Hakuba Jumping Stadium due to neglect of the deck part
of the landing slope in the model calculation. The calcu-
lation result is that Dv is roughly in proportion to U0 but the
inclination is only slightly larger at large U0: for example,
83 m for U0 = 23 m/s (=82.8 km/h), 103.5 m for
U0 = 26 m/s (=93.6 km/h).
3.2.2 Correlation results: a - Dr, c - Dr, h - Dr
and tf - Dr correlations
The results of a - Dr, c - Dr, h - Dr and tf - Dr corre-
lations are shown respectively in Fig. 3a–d, and the data of
the coefficient of determination for the linear regressions,
R2, are shown in Table 1. The values of a, c, h and tf were
Fig. 2 Time variations of the lift area SL and some angles of a jumper
posture for two typical cases (an advanced jumper and a beginner
jumper). Data are numerically smoothed out without applying
numerical low-pass filtering: a for the lift area SL and the ski angle
of attack a, b for the body angle of attack c, and c for the body
forward leaning angle h
M. Murakami et al.
measured at 1.4 s after takeoff, when every flight has been
about to enter a quasi-steady flight phase. The result sug-
gests that the angles a, c and h should be smaller for larger
flight distance. It is because smaller a, c and h lead to
smaller aerodynamic drag. It should be noted that the
relation c = h ? a holds as seen from the definition shown
in Fig. 1. This relation indicates that only two among c, hand a are independent. A typical jumping posture of
beginners is that with small h but large a, that is the posture
with small ski angle of attack but an almost upright body
position. In this case, c inevitably becomes large, and thus
Dr is not large even if a is small. In the case of large a and cin the early jumping flight phase, a jumper flies while
possibly undergoing the effect of aerodynamic stall, char-
acterized by large drag force. The angle h that is easily
quantified even from a still picture tends to be regarded as
being important. However, it is the parameters a and cdefined with respect to the vector U of the airflow around a
jumper, or equivalently of the jumper flight path, that are
decisive of the aerodynamic force exerted on a jumper. The
flight posture having small a and c, and thus small h, results
in longer flight distance. The posture having small h, which
is sometimes considered as a target of the skill that
beginners should acquire, is effective only if a is small. The
result of Fig. 3d indicates that quick completion of initial
movement subsequent to takeoff is important for larger Dj.
Immediately after takeoff, jumpers nearly stand by upright
posture against airflow resulting in larger h causing larger
drag. Therefore, quick transition to a quasi-steady flight is
desirable for a long flight distance.
It is interesting to compare the data between the
advanced and beginner jumpers. It is evident in Fig. 3 that
every parameter, a, c, h and tf for advanced jumpers, is, on
average, smaller than for beginners. The flight posture with
small a, c, h and tf assures the condition of a flight posture
with small drag. In particular for the parameters, c and h, of
advanced jumpers the data points are rather concentrated in
narrower and more relevant regime for larger Dr. Thus, the
correlations of these parameters with Dr are quite low.
Smaller data scattering of these parameters for advanced
Fig. 3 Correlation results. a for
a - Dr correlation, b for
c - Dr correlation, c for h - Dr
correlation and d for tf - Dr
correlation. The solid line is a
linear least square regression
line for all the data
Table 1 Coefficient of
determination for linear
regression results
Regression a - Dr c - Dr H - Dr tf - Dr f - c f - h f - Dr
Figure 3a 3b 3c 3d 5a 5b 5c
Coefficient of determination R2 0.5482 0.8376 0.7489 0.4691 0.7436 0.7092 0.7304
Image analysis of ski jumping flight posture
jumpers suggests that the flights are rather reproducible,
and thus, they can perform two good jumps in high prob-
ability in a jumping competition. Data scattering is possi-
bly attributed to the contribution from some uncontrolled
factors, such as wind, the direction and timing of takeoff,
unnecessary movement of arms, etc. It may be considered
that the jumping flight of beginners is far more affected by
such uncertain factors. The correlations of parameters, cand h, with Dr for beginners are higher than those of
advanced jumpers. The regression lines seem to indicate
the path along which the jump skill of beginners is
improved. Beginners may ultimately reach the data area of
advanced jumpers that exists on the natural extension of
each regression line. On the other hand, the data distribu-
tions of a and tf are rather multi-layered in data group
structures between advanced and beginner jumpers. The
data of advanced jumpers are always on the higher side of
Dr than those of beginners at the same a or tf value. That is,
if beginners are continuing training of the beginner level
aimlessly, they get used only to a skillful beginner at most.
To reach the level of advanced jumpers, making a highly
motivated training such as the leap of something as
described below appears to be necessary. It is suggested
that it is insufficient for large Dj for beginners to have
mastered to only make a or tf small. It is quite true that
acquisition of the skill for a jumping posture with small
body angle of attack c is a key for beginner jumpers to
progress to a higher step. However, two points should be
paid attention to for their training. The first is that fear is
followed in order to take a jumping posture with small cbecause the transition from takeoff to this flight posture
requires throwing the trunk forward. The second is that it is
necessary to understand that the nose-down moment
(pitching moment in the direction where the head is going
down) must be given to the trunk as an initial condition of
flight dynamics at the instant of takeoff to achieve a
jumping posture with small c, because the nose-down
moment produced aerodynamically in the air during a
jumping flight is quite small.
3.2.3 Effect of the angle between CG-boots line and ski
at takeoff
As pointed out in the last section, for small c, which is one
of the keys for large Dj, nose-down moment should be
given to jumper’s trunk in the takeoff motion. However,
the direct measurement of the pitching moment from the
high-speed video images could not be performed because
the nose-down rotation of a jumper trunk immediately
before the takeoff point is not visible because a jumper was
screened by the opaque safety board. Instead, we selected
the angle between the ski (or the slope line at the takeoff
point) and the CG-boots line of a jumper immediately after
takeoff, f, as a plausible measure of the nose-down
moment at the instant of takeoff. The definition of f is
illustrated in Fig. 4. The angle f was measured from the
images taken with the auxiliary camera. The correlations
are shown in Fig. 5a–c respectively for f - c, f - h and
f - Dr correlations, and the data of the coefficient of
determination, R2, for the linear regressions are shown in
Table 1. It is seen that there are strong correlations among
them. This result means smaller f is a condition for sub-
sequent smaller c. This may be the reason why f is strongly
correlated with Dr. This may be regarded as the evidence
that f plays a role as the measure of the nose-down moment
at the instant of takeoff. The data of advanced jumpers
exist in a narrower regime on the extension of corre-
sponding regression lines of beginners.
3.3 Consideration on the effects of aerodynamic forces
The correlations between the drag area SD and Dr, and the
lift area SL and Dr, shown in Fig. 6a and b respectively, are
examined to investigate the aerodynamic force effects. The
detailed statistical data of the linear regressions are pre-
sented in Table 2. It is seen that there are a number of
definite differences in these two functional relations
between the data of the advanced and the beginner jump-
ers. For the data of the advanced jumpers, there are almost
no correlations in both SD - Dr and SL - Dr relations in
the following two different aspects: As for SD, the data
points are distributed in a narrow region of the SD - Dr
plane. However for SL, the data show rather wide distri-
bution in SL, but Dr hardly depends on SL, and the data
points for the advanced jumpers are on the larger side of Dr
than those of the beginner jumpers. The data point distri-
bution in the SD - Dr diagram in Fig. 6a indicates that the
data for the advanced jumpers are around the extension line
Fig. 4 Definition of the angle f between the ski and the CG-boots
line
M. Murakami et al.
of the regression for the beginners. This seems to suggest
that beginners can arrive at the level of advanced jumpers,
if ordinary efforts are continued. It is, on the other hand,
seen in the SL - Dr diagram that the data of the advanced
jumpers are located above the data point row of the
beginner jumpers. This consequence may suggest that
beginners cannot arrive at the level of advanced jumpers
after just another continuation of training. Some special
device and effort for higher skill are required for beginners
to grow up to be advanced jumpers. Mastery of a jumping
posture with small body angle of attack c for beginner
jumpers as described in the Sect. 3.2.2 may be in line with
this. It is fair to conclude that there may be weak corre-
lations in SD - Dr and SL - Dr relations if the data of the
advanced and the beginner jumpers are taken into account
all together. This result seems to suggest that the posture
with the smaller SD and SL makes Dr the larger. If this
supposition were true, a crouching jumping posture would
be the best flight posture as it fulfills the condition of small
SD and small SL simultaneously. The requirement of small
SD is quite reasonable for larger Dr. But it does not seem a
guiding principle for larger Dr and rather seems even
unreasonable that the smaller SL is the better. This seem-
ingly unreasonable consequence originates from the vari-
ations of SD and SL as a function of ski angle of attack ashown in Fig. 7, where small SD inevitably makes SL small
except in the aerodynamic stall region. It is, on the other
hand, a plain fact of aerodynamics that SD should be small
and SL should be large for larger flight distance Dj. The best
performance would certainly be achieved at some inter-
mediate values of SD and SL.
It is, then, investigated how much values of SD and SL
should be for the maximum jumping flight distance. It is a
well-known fact in aerodynamics that the maximum L/D,
or equivalently the maximum SL/SD, results in the maxi-
mum flight distance as long as it is in a steady flight. In
Fig. 8 the correlation between SL/SD and Dr is shown, and
the statistical data of linear regression are shown in
Table 2. This result shows the correlation is significant
only for the beginners, where a weakly positive correlation
is recognized, while for the advanced jumpers there is
Fig. 5 Correlation results. a For f - c correlation, b for f - h correlation and c for f - Dr correlation. The solid line is a linear least square
regression line for all the data
Fig. 6 Correlation results. a for SD - Dr correlation and b for SL -
Dr correlation. The solid line is a linear least square regression line for
all the data
Image analysis of ski jumping flight posture
almost no correlation between them, and merely SL/SD
value is larger than 0.95. The result, in fact, indicates that
the maximum SL/SD does not always give the result of
maximum jumping flight distance. Of course, all the
jumping flights examined here being in the early flight
phase were never in a steady state. For further consider-
ation, the present data are plotted on a SD - SL polar
diagram as given in Fig. 9, where the data are classified
according to the magnitude of Dr. In this figure the data
presented in Fig. 7 are also plotted in the form of a polar
curve just as an example for the case of a jumping posture
of straight trunk without a bend of the waist with h = 10�
and a ski opening angle k of 25�, though this jumping
posture is not a typical one in the early flight phase. It is
seen in the result that the value of SD never exceeds 0.6 for
the advanced jumpers, as also evident from the result of
Fig. 6a. As a matter of fact, this result indicates one of the
decisive differences between the advanced and the begin-
ner jumpers as follows. For the advanced jumpers, SL/SD is
larger than 0.95, though it is not a definite guiding principle
for larger flight distance that the larger SL/SD is the better.
In other words, small SD is desirable, but too small SD is
not recommended as this condition makes SL also too
small. Some lift force is certainly necessary, which, toge-
ther with the condition that SD is small, is reflected in the
condition of SL/SD [ 0.95 for satisfactory jumping flight
distance. The straight trunk posture with too deep forward
leaning is not so good for longer jumping flight distance
because of small SL though SD is small. Furthermore, this
Table 2 Linear regression statistics on aerodynamic data
SD - Dr SL - Dr SL/SD - Dr
Figure 6a 6b 6c
All data
Coefficient of determination R2 0.5353 0.1962 0.3348
p level 2.16E-5 2.34E-2 1.96E-3
H0 (5 %) rejected Yesa Yes Yes
Advanced jumpers
Coefficient of determination R2 0.0121 5.88E-3 2.32E-3
p level 0.733 0.813 0.882
H0 (5 %) rejected Nob No No
Beginner jumpers
Coefficient of determination R2 0.5628 0.232 0.524
p level 3.14E-3 8.10E-2 3.40E-3
H0 (5 %) rejected Yes No Yes
Yesa: H0 (Dr is independent of SD) is rejected
Nob: H0 (Dr is independent of SD) is not rejected
Fig. 7 An example of the wind tunnel measurement data of SD and SL
plotted against a [5]. The forward leaning angle h is 10� and the ski
opening angle k is 25�
Fig. 8 SL/SD - Dr correlation result. The solid line is a linear least
square regression line for all the data
Fig. 9 Flight experimental data plotted on a SL - SD polar diagram.
The data are classified according to the magnitude of Dr. The straight
lines are those of constant SL/SD, and the solid curve is the polar curve
of the data presented in Fig. 7
M. Murakami et al.
posture is disadvantageous in a quasi-steady flight phase
due to small SL because the reattachment of separated flow
is hard to occur. It should be noted that all the flight test
data of Fig. 9 are aerodynamically inferior to the wind
tunnel data of Fig. 7 in the sense that the flight experi-
mental data points are all below the polar curve of the data
of Fig. 7. It is primarily because the forward leaning angles
h of the present flight data are larger than 10� and the data
were taken in the early flight phase before reaching a quasi-
steady flight phase.
4 Conclusions
The conclusions of this study are summarized as follows:
• The body angle of attack c, the ski angle of attack a, the
body forward leaning angle h, and the transient time to
a quasi-steady flight tf should all be small for larger
jumping flight distance, which is a necessary condition
for small drag force, in particular in the early flight
phase. Among these parameters, c is the most
influential.
• The distribution of these measurement data is clearly
distinguished between advanced jumpers and beginner
jumpers. The data of advanced jumpers are found in the
small value ranges of a, c, h and tf, and furthermore the
data scattering of advanced jumpers is smaller than
those of beginners. Typical fault of beginner jumpers is
too large c and a resulting in too large drag caused by
aerodynamic stall.
• In the early flight phase, a flight posture with small drag
is a key for long flight distance with the condition that
SL/SD is larger than 1.0.
Acknowledgments This research was financially supported by the
Grant-in-Aid for Scientific Research from the Japan Society for the
Promotion of Science (19650167).
References
1. Tani I, Mitsuishi T (1951) Aerodynamics of ski jumping. Science
(Japanese edition) 117–152. (in Japanese)
2. Tani I, Iuchi M (1971) Flight-mechanical investigation of ski
jumping. Scientific study of skiing in Japan. Hitachi 35–53
3. Jin H, Shimizu S, Watanuki T, Kubota H, Kobayashi K (1995)
Desirable gliding style and technique in ski jumping. J Appl
Biomech 11:460–474
4. Muller W, Platzer D, Schmolzer B (1996) Dynamics of human
flight on skis: improvements in safety and fairness in ski jumping.
J Biomech 29–8:1061–1068
5. Seo K, Watanabe I, Murakami M (2004) Aerodynamic force data
for a V-style ski jumping flight. Sports Eng 7–1:1–39
6. Arndt A, Bruggemann GP, Virmavirta M, Komi P (1995) Tech-
niques used by Olympia ski jumpers in the transition from takeoff
to early flight. J Appl Biomech 11:224–237
7. Hubbard M, Hibbard RL, Yeadon MR, Komor A (1989) A
multisegment dynamic model of ski jumping. Int J Sport Bio-
mech 5:258–274
8. Seo K, Murakami M, Yoshida K (2004) Optimal flight technique
for V-style ski jumping. Sports Eng 7–2:97–104
9. Murakami M, Hirai N, Seo K, Ohgi Y (2007) Aerodynamic
forces computation from high-speed video image of ski jumping
flight. In: Proceedings of Asia-Pacific Congress on Sports Tech-
nology pp 857–861
10. Murakami M, Hirai N, Seo K, Ohgi Y (2008) Aerodynamic study
of ski jumping flight based on high-speed video image. In: Pro-
ceedings of 7th ISEA, the Engineering of sport vol 7, pp 449–456
11. Chandler RF (1975) Investigation of inertial properties of the
human body. Technical Report AMRL-74-137, Wright Patterson
Air Force Base
Image analysis of ski jumping flight posture