3

从视频记录中,可以提取行人头部穿过瓶颈的轨迹。因为,行人在移动轨迹时摇摆包含我想消除的讨厌的振荡。(平滑算法与这个问题无关)

结果如图所示

所以,我的目标是使用蓝点(称为零点)获得平滑的轨迹。为此,我想到了样条曲线。不幸的是,scipy 的函数splrep有一个限制,即数据应该相对于 x 轴进行排序(为什么?)。

但是在我的情况下,对数据点进行排序不是一个好主意,然后它会使真实轨迹变形。样条曲线的结果,排序后看起来

(图的最后部分是第二部分的放大图,以展示振荡。)

从数学的角度来看,splrep的结果可能是正确的,但是它们没有物理相关性,这是因为我不必操纵轨迹。

任何想法如何用python(scipy)正确解决这个问题?

如果有人想玩数据,这是我的代码。

from matplotlib.pyplot import *
from numpy import *
from scipy import interpolate

# trajectory
path =   array([
[ 3.83911987e+02,  -3.63506010e+01],
[ 3.80407013e+02,  -3.70444980e+01],
[ 3.77910004e+02,  -3.73224980e+01],
[ 3.75592010e+02,  -3.72258990e+01],
[ 3.72606995e+02,  -3.68973010e+01],
[ 3.68860992e+02,  -3.64117010e+01],
[ 3.64709991e+02,  -3.54816020e+01],
[ 3.60441986e+02,  -3.46052020e+01],
[ 3.56470001e+02,  -3.33995020e+01],
[ 3.53148010e+02,  -3.17936990e+01],
[ 3.50332001e+02,  -2.99004990e+01],
[ 3.48214996e+02,  -2.76462990e+01],
[ 3.45072998e+02,  -2.56490000e+01],
[ 3.41787994e+02,  -2.45536990e+01],
[ 3.38785004e+02,  -2.41054990e+01],
[ 3.36031006e+02,  -2.40007990e+01],
[ 3.33045013e+02,  -2.41030010e+01],
[ 3.29737000e+02,  -2.44983010e+01],
[ 3.26299011e+02,  -2.53087010e+01],
[ 3.22807007e+02,  -2.65634990e+01],
[ 3.19296997e+02,  -2.82255000e+01],
[ 3.15854004e+02,  -3.02551990e+01],
[ 3.12652008e+02,  -3.22993010e+01],
[ 3.09871002e+02,  -3.41482010e+01],
[ 3.07183990e+02,  -3.56874010e+01],
[ 3.04494995e+02,  -3.66516990e+01],
[ 3.01890015e+02,  -3.71099010e+01],
[ 2.99341003e+02,  -3.74341010e+01],
[ 2.96681000e+02,  -3.77743990e+01],
[ 2.93803986e+02,  -3.79683000e+01],
[ 2.90785004e+02,  -3.77564010e+01],
[ 2.87817993e+02,  -3.69802020e+01],
[ 2.84980988e+02,  -3.57428020e+01],
[ 2.82131989e+02,  -3.40877990e+01],
[ 2.79335999e+02,  -3.21539990e+01],
[ 2.76933014e+02,  -3.00962010e+01],
[ 2.74431000e+02,  -2.81432000e+01],
[ 2.71325012e+02,  -2.63622000e+01],
[ 2.68205994e+02,  -2.47146000e+01],
[ 2.65588013e+02,  -2.33606000e+01],
[ 2.63167999e+02,  -2.23937000e+01],
[ 2.60816010e+02,  -2.16775000e+01],
[ 2.58609985e+02,  -2.10203990e+01],
[ 2.56585999e+02,  -2.06432990e+01],
[ 2.54651001e+02,  -2.06508010e+01],
[ 2.52639008e+02,  -2.09992010e+01],
[ 2.50608002e+02,  -2.17117000e+01],
[ 2.48576004e+02,  -2.28300990e+01],
[ 2.46585999e+02,  -2.42358000e+01],
[ 2.44641006e+02,  -2.59650990e+01],
[ 2.42638000e+02,  -2.79718000e+01],
[ 2.40481995e+02,  -2.97957990e+01],
[ 2.38220993e+02,  -3.15263000e+01],
[ 2.35929993e+02,  -3.32879980e+01],
[ 2.33899994e+02,  -3.47654000e+01],
[ 2.32324005e+02,  -3.58550000e+01],
[ 2.30934006e+02,  -3.66335980e+01],
[ 2.29505005e+02,  -3.73493000e+01],
[ 2.28022995e+02,  -3.80163000e+01],
[ 2.26539993e+02,  -3.84826010e+01],
[ 2.25024002e+02,  -3.86341020e+01],
[ 2.23253998e+02,  -3.84396020e+01],
[ 2.21259995e+02,  -3.78853990e+01],
[ 2.19130997e+02,  -3.69795990e+01],
[ 2.16889999e+02,  -3.58032990e+01],
[ 2.14451996e+02,  -3.43969990e+01],
[ 2.11820007e+02,  -3.27537000e+01],
[ 2.09186996e+02,  -3.10618990e+01],
[ 2.07384995e+02,  -2.93244990e+01],
[ 2.04218002e+02,  -2.81264000e+01],
[ 2.01115997e+02,  -2.68684010e+01],
[ 1.98009003e+02,  -2.56004010e+01],
[ 1.95281006e+02,  -2.47794000e+01],
[ 1.92839005e+02,  -2.42327000e+01],
[ 1.90753006e+02,  -2.39883000e+01],
[ 1.88904007e+02,  -2.39496990e+01],
[ 1.87065002e+02,  -2.40893000e+01],
[ 1.85164001e+02,  -2.51790010e+01],
[ 1.83408005e+02,  -2.58622000e+01],
[ 1.81880997e+02,  -2.68085000e+01],
[ 1.80416000e+02,  -2.82280010e+01],
[ 1.79287994e+02,  -3.01436000e+01],
[ 1.78574997e+02,  -3.22188990e+01],
[ 1.77757004e+02,  -3.42841000e+01],
[ 1.76931000e+02,  -3.64440990e+01],
[ 1.76029007e+02,  -3.81259990e+01],
[ 1.75113998e+02,  -3.96713980e+01],
[ 1.74244003e+02,  -4.08652000e+01],
[ 1.73044998e+02,  -4.15327990e+01],
[ 1.72110001e+02,  -4.16768000e+01],
[ 1.70936005e+02,  -4.17703020e+01],
[ 1.69544006e+02,  -4.12965010e+01],
[ 1.67804993e+02,  -4.05098000e+01],
[ 1.66028000e+02,  -3.96259000e+01],
[ 1.64475998e+02,  -3.81104010e+01],
[ 1.63324005e+02,  -3.61517980e+01],
[ 1.62007004e+02,  -3.43326990e+01],
[ 1.60423996e+02,  -3.27486990e+01],
[ 1.58707993e+02,  -3.09353010e+01],
[ 1.56770996e+02,  -2.94423010e+01],
[ 1.54835999e+02,  -2.79886000e+01],
[ 1.52953003e+02,  -2.67169000e+01],
[ 1.50912994e+02,  -2.57530990e+01],
[ 1.48996002e+02,  -2.54090000e+01],
[ 1.47061996e+02,  -2.53249000e+01],
[ 1.45320007e+02,  -2.55256000e+01],
[ 1.43707993e+02,  -2.60972000e+01],
[ 1.41876007e+02,  -2.68913990e+01],
[ 1.40018997e+02,  -2.78764000e+01],
[ 1.38016998e+02,  -2.91917000e+01],
[ 1.36063004e+02,  -3.06130010e+01],
[ 1.34001999e+02,  -3.22636990e+01],
[ 1.31807999e+02,  -3.42985000e+01],
[ 1.29455002e+02,  -3.65051990e+01],
[ 1.27331001e+02,  -3.85568010e+01],
[ 1.25116997e+02,  -4.05253980e+01],
[ 1.23258003e+02,  -4.22365990e+01],
[ 1.21709000e+02,  -4.36623000e+01],
[ 1.20507004e+02,  -4.48983990e+01],
[ 1.19530998e+02,  -4.56188010e+01],
[ 1.18888000e+02,  -4.60003010e+01],
[ 1.18030998e+02,  -4.58804020e+01],
[ 1.17032997e+02,  -4.53919980e+01],
[ 1.15643997e+02,  -4.47540020e+01],
[ 1.14077003e+02,  -4.39210010e+01],
[ 1.12264000e+02,  -4.29716000e+01],
[ 1.10446999e+02,  -4.18083990e+01],
[ 1.08487000e+02,  -4.06045990e+01],
[ 1.06468002e+02,  -3.88120990e+01],
[ 1.04453003e+02,  -3.67605020e+01],
[ 1.02263000e+02,  -3.47210010e+01],
[ 9.98833010e+01,  -3.29178010e+01],
[ 9.76293030e+01,  -3.16152000e+01],
[ 9.55121000e+01,  -3.09032000e+01],
[ 9.33936000e+01,  -3.05140000e+01],
[ 9.14021000e+01,  -3.02255000e+01],
[ 8.95216980e+01,  -3.02507990e+01],
[ 8.76836010e+01,  -3.06539000e+01],
[ 8.58259960e+01,  -3.10028990e+01],
[ 8.40342030e+01,  -3.14128000e+01],
[ 8.25109020e+01,  -3.24874990e+01],
[ 8.16477970e+01,  -3.37783010e+01],
[ 8.10721970e+01,  -3.54217000e+01],
[ 8.02864000e+01,  -3.69583020e+01],
[ 7.93555980e+01,  -3.81934010e+01],
[ 7.82388990e+01,  -3.90952990e+01],
[ 7.72057040e+01,  -3.95537990e+01],
[ 7.61997990e+01,  -3.97468990e+01],
[ 7.54201970e+01,  -3.95574000e+01],
[ 7.46388020e+01,  -3.92573010e+01],
[ 7.36848980e+01,  -3.86537020e+01],
[ 7.25225980e+01,  -3.76381990e+01],
[ 7.13074040e+01,  -3.64692990e+01],
[ 7.02815020e+01,  -3.50397990e+01],
[ 6.93098980e+01,  -3.34804990e+01],
[ 6.82506030e+01,  -3.18008000e+01],
[ 6.69611970e+01,  -3.02639010e+01],
[ 6.55927960e+01,  -2.89077000e+01],
[ 6.44023970e+01,  -2.77817990e+01],
[ 6.31432990e+01,  -2.70891990e+01],
[ 6.19845010e+01,  -2.66151010e+01],
[ 6.08012010e+01,  -2.63731990e+01],
[ 5.95279010e+01,  -2.63244000e+01],
[ 5.81848980e+01,  -2.66625000e+01],
[ 5.67341000e+01,  -2.70667000e+01],
[ 5.53522000e+01,  -2.76193010e+01],
[ 5.40955010e+01,  -2.89263990e+01],
[ 5.30055010e+01,  -3.03295990e+01],
[ 5.19216000e+01,  -3.15974010e+01],
[ 5.08109020e+01,  -3.27397990e+01],
[ 4.95959010e+01,  -3.36008990e+01],
[ 4.83475990e+01,  -3.42673000e+01],
[ 4.69407010e+01,  -3.46990010e+01],
[ 4.56283000e+01,  -3.50008010e+01],
[ 4.45952990e+01,  -3.53820990e+01],
[ 4.35136990e+01,  -3.55634990e+01],
[ 4.24670980e+01,  -3.54066010e+01],
[ 4.11960980e+01,  -3.49855000e+01],
[ 3.94818000e+01,  -3.43259010e+01],
[ 3.75299990e+01,  -3.32298010e+01],
[ 3.54291000e+01,  -3.18832000e+01],
[ 3.35419010e+01,  -3.01863990e+01],
[ 3.18060000e+01,  -2.83223990e+01],
[ 3.03379000e+01,  -2.64284990e+01],
[ 2.90386010e+01,  -2.48046000e+01],
[ 2.77717000e+01,  -2.35111010e+01],
[ 2.66282010e+01,  -2.24167000e+01],
[ 2.56504000e+01,  -2.15105990e+01],
[ 2.45620990e+01,  -2.09494990e+01],
[ 2.36200010e+01,  -2.05372010e+01],
[ 2.25006010e+01,  -2.02623000e+01],
[ 2.12878000e+01,  -2.01607000e+01],
[ 1.99335990e+01,  -2.02614990e+01],
[ 1.84620000e+01,  -2.04685990e+01],
[ 1.67920000e+01,  -2.09810010e+01],
[ 1.50698000e+01,  -2.15951000e+01],
[ 1.31372000e+01,  -2.24368000e+01],
[ 1.08794000e+01,  -2.30415000e+01],
[ 8.29891000e+00,  -2.31980000e+01],
[ 5.61593000e+00,  -2.27952000e+01],
[ 2.89633000e+00,  -2.16718010e+01],
[ 8.38298000e-01,  -2.05924000e+01],
[-9.50327000e-01,  -1.92878000e+01],
[-2.68694000e+00,  -1.77108000e+01],
[-4.20826000e+00,  -1.59780000e+01],
[-5.74339000e+00,  -1.44248000e+01],
[-7.13617000e+00,  -1.34790000e+01],
[-8.80355000e+00,  -1.29467000e+01],
[-1.09091000e+01,  -1.23274000e+01],
[-1.32822000e+01,  -1.15982000e+01],
[-1.57716000e+01,  -1.08766000e+01],
[-1.82273010e+01,  -1.04894000e+01],
[-2.05464000e+01,  -1.03020000e+01],
[-2.24237000e+01,  -9.95550000e+00],
[-2.38675990e+01,  -9.45586000e+00],
[-2.50128990e+01,  -8.74390000e+00],
[-2.58752990e+01,  -7.79803000e+00],
[-2.64664000e+01,  -6.62279000e+00],
[-2.66699010e+01,  -5.05463000e+00],
[-2.65776000e+01,  -3.41969000e+00],
[-2.63474010e+01,  -1.57144000e+00],
[-2.61133000e+01,   6.46999000e-01],
[-2.58085990e+01,   3.23445000e+00],
[-2.55969010e+01,   5.78099000e+00],
[-2.55000000e+01,   8.36744000e+00],
[-2.54680000e+01,   1.11505000e+01],
[-2.53899000e+01,   1.41447000e+01],
[-2.53566000e+01,   1.71451000e+01],
[-2.58187010e+01,   2.05758000e+01],
[-2.65830990e+01,   2.44065000e+01],
[-2.79578990e+01,   2.84622990e+01],
[-2.95126990e+01,   3.26220020e+01],
[-3.12829000e+01,   3.70005990e+01],
[-3.31349980e+01,   4.12872010e+01],
[-3.51132010e+01,   4.54515000e+01],
[-3.70420000e+01,   4.90353010e+01],
[-3.85575980e+01,   5.27621990e+01],
[-3.94586980e+01,   5.64221990e+01],
[-3.97930980e+01,   5.98333020e+01],
[-3.98658980e+01,   6.29235000e+01],
[-3.94589000e+01,   6.64207990e+01],
[-3.88973010e+01,   6.96651990e+01],
[-3.81236990e+01,   7.34041980e+01],
[-3.68148000e+01,   7.73573990e+01],
[-3.54217990e+01,   8.15375980e+01],
[-3.38608020e+01,   8.63582990e+01],
[-3.15564000e+01,   9.19246980e+01],
[-2.93428990e+01,   9.76631010e+01],
[-2.72516990e+01,   1.02860001e+02],
[-2.59647010e+01,   1.08068001e+02],
[-2.51345010e+01,   1.12709999e+02],
[-2.48617990e+01,   1.16903000e+02],
[-2.48806990e+01,   1.20731003e+02],
[-2.52887990e+01,   1.24375000e+02],
[-2.62332990e+01,   1.28048004e+02],
[-2.74224000e+01,   1.31630005e+02],
[-2.93577000e+01,   1.35138000e+02],
[-3.18421000e+01,   1.38580002e+02],
[-3.43532980e+01,   1.42533997e+02],
[-3.68244020e+01,   1.46210999e+02],
[-3.83947980e+01,   1.49451996e+02],
[-3.94422990e+01,   1.52266006e+02],
[-4.00191000e+01,   1.54524002e+02],
[-4.01052020e+01,   1.56955002e+02],
[-4.01661990e+01,   1.59701996e+02],
[-3.94771000e+01,   1.62824997e+02],
[-3.84656980e+01,   1.65957993e+02],
[-3.69557000e+01,   1.69384995e+02],
[-3.54382020e+01,   1.72932007e+02],
[-3.37109990e+01,   1.77065002e+02],
[-3.12064990e+01,   1.81628006e+02],
[-2.85994000e+01,   1.86921005e+02],
[-2.57090000e+01,   1.91203003e+02],
[-2.33045010e+01,   1.95505005e+02],
[-2.18305000e+01,   1.99636993e+02],
[-2.07416990e+01,   2.03414993e+02],
[-2.00366990e+01,   2.06847000e+02],
[-1.97754000e+01,   2.11009003e+02],
[-1.97579000e+01,   2.15139999e+02],
[-1.99501990e+01,   2.19901001e+02],
[-2.07133010e+01,   2.24968994e+02],
[-2.17415010e+01,   2.30507004e+02],
[-2.29566990e+01,   2.36494003e+02],
[-2.40505010e+01,   2.42442001e+02],
[-2.48687990e+01,   2.48417007e+02],
[-2.52953000e+01,   2.54072006e+02],
[-2.54123000e+01,   2.59537994e+02],
[-2.54193000e+01,   2.64915985e+02],
[-2.56805990e+01,   2.71098999e+02],
[-2.59018990e+01,   2.77950989e+02],
[-2.64182000e+01,   2.84119995e+02],
[-2.65342010e+01,   2.89997009e+02],
[-2.60492990e+01,   2.95882996e+02],
[-2.51636010e+01,   3.02220001e+02],
[-2.36994990e+01,   3.09553986e+02],
[-2.22290000e+01,   3.16851990e+02],
[-2.17989010e+01,   3.23812012e+02],
[-2.21809010e+01,   3.30109985e+02],
[-2.24906010e+01,   3.36039001e+02],
[-2.24252000e+01,   3.41493011e+02],
[-2.26278990e+01,   3.46933990e+02],
[-2.34319000e+01,   3.52662994e+02],
[-2.43871990e+01,   3.58191010e+02],
[-2.52982010e+01,   3.64041992e+02],
[-2.64419000e+01,   3.70739990e+02],
[-2.79559990e+01,   3.77709991e+02],
[-2.87983000e+01,   3.84595001e+02],
[-2.86500000e+01,   3.90970001e+02],
[-2.82661990e+01,   3.96442993e+02],
[-2.81972010e+01,   4.01808014e+02]])   
#nullpoints
A = array([
[ 3.50332001e+02,  -2.99004990e+01],
[ 3.15854004e+02,  -3.02551990e+01],
[ 2.76933014e+02,  -3.00962010e+01],
[ 2.42638000e+02,  -2.79718000e+01],
[ 2.11820007e+02,  -3.27537000e+01],
[ 1.77757004e+02,  -3.42841000e+01],
[ 1.63324005e+02,  -3.61517980e+01],
[ 1.27331001e+02,  -3.85568010e+01],
[ 1.06468002e+02,  -3.88120990e+01],
[ 8.10721970e+01,  -3.54217000e+01],
[ 7.02815020e+01,  -3.50397990e+01],
[ 5.30055010e+01,  -3.03295990e+01],
[ 3.03379000e+01,  -2.64284990e+01],
[ 1.67920000e+01,  -2.09810010e+01],
[-2.61133000e+01,   6.46999000e-01],
[-3.31349980e+01,   4.12872010e+01],
[-3.38608020e+01,   8.63582990e+01],
[-3.18421000e+01,   1.38580002e+02],
[-3.12064990e+01,   1.81628006e+02],
[-2.17415010e+01,   2.30507004e+02],
[-2.17415010e+01,   2.30507004e+02],
[-2.51636010e+01,   3.02220001e+02],
[-2.43871990e+01,   3.58191010e+02]])

isSort = 1 
# isSort = 0 --> problems with splrep 
#    File "/usr/local/lib/python2.6/dist-packages/scipy/interpolate/fitpack.py",
#    line 466, in splrep
#    raise _iermess[ier][3](_iermess[ier][0])
#    ValueError:     Error on input data

if isSort:
    I = lexsort((A[:,1], A[:,0]) )
    A = A[I]

a1 = A[:,0]
a2 = A[:,1] 
subplot(311)
plot(path[:,0], path[:,1], "-r", lw=2, label="path")
plot(a1, a2, "ob", ms = 6,  label = "nullpoints")
legend();grid()
subplot(312)
# -------------------- spline ------------------------
xnew = np.arange( min(a1), max(a1), 1)
for s, color in zip([abs(max(a2)- min(a2))*32, abs(max(a2)- min(a2))*8], ["-m", "-g"]):
    tck = interpolate.splrep(a1, a2, s = s)
    ynew = interpolate.splev(xnew, tck)
    plot(xnew, ynew, "%s"%color, lw=2, label = "splrep, s=%d"%s)
#------------------------------------------------------
plot(a1, a2, "ob", ms = 6)
legend();grid()
# zoom around the oscillating part
subplot(313)
for s, color in zip([abs(max(a2)- min(a2))*32, abs(max(a2)- min(a2))*8], ["-m", "-g"]):
    tck = interpolate.splrep(a1, a2, s = s)
    ynew = interpolate.splev(xnew, tck)
    plot(xnew, ynew, "%s"%color, lw=2, label = "splrep, s=%d"%s)
xlim([min(a2), -20])
plot(a1, a2, "ob", ms = 6,  label = "nullpoints")
grid();savefig("dummy.png"); show()
4

1 回答 1

5

您的数据的问题似乎是重复的:当两个连续点相等时,您无法正确插值。

在“空点”数据的末尾附近,有两个相同的行:

[-2.17415010e+01, 2.30507004e+02],
[-2.17415010e+01, 2.30507004e+02],

另请注意,由于您有 2D 数据,因此您需要使用splprep()(注意额外的p)。

对于我的一个项目,我使用以下函数来插入 2D 点集(折线):

import numpy as np
import scipy.interpolate as interp

def interpolate_polyline(polyline, num_points):
    duplicates = []
    for i in range(1, len(polyline)):
        if np.allclose(polyline[i], polyline[i-1]):
            duplicates.append(i)
    if duplicates:
        polyline = np.delete(polyline, duplicates, axis=0)
    tck, u = interp.splprep(polyline.T, s=0)
    u = np.linspace(0.0, 1.0, num_points)
    return np.column_stack(interp.splev(u, tck))

我可以成功地将此功能应用于您的“空点”数据:

>>> B = interpolate_polyline(A, 100)
>>> B.shape
(100, 2)
>>> plot(B[:, 0], B[:, 1])

情节结果

于 2012-11-22T13:22:50.490 回答