ARUCO markers: basics¶

1: Marker creation¶

import numpy as np
import cv2, PIL
from cv2 import aruco
import matplotlib.pyplot as plt
import matplotlib as mpl
import pandas as pd
%matplotlib nbagg

aruco_dict = aruco.Dictionary_get(aruco.DICT_6X6_250)

fig = plt.figure()
nx = 4
ny = 3
for i in range(1, nx*ny+1):
    ax = fig.add_subplot(ny,nx, i)
    img = aruco.drawMarker(aruco_dict,i, 700)
    plt.imshow(img, cmap = mpl.cm.gray, interpolation = "nearest")
    ax.axis("off")

plt.savefig("_data/markers.jpeg")
plt.show()

<IPython.core.display.Javascript object>

2: Print, cut, stick and take a picture¶

frame = cv2.imread("_data/marqueurs_chaise.jpg")
plt.figure()
plt.imshow(frame)
plt.show()

<IPython.core.display.Javascript object>

3: Post processing¶

%%time
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
aruco_dict = aruco.Dictionary_get(aruco.DICT_6X6_250)
parameters =  aruco.DetectorParameters_create()
corners, ids, rejectedImgPoints = aruco.detectMarkers(gray, aruco_dict, parameters=parameters)
frame_markers = aruco.drawDetectedMarkers(frame.copy(), corners, ids)

Wall time: 178 ms

rejectedImgPoints[1]

array([[[ 1213.,  1229.],
        [ 1217.,  1221.],
        [ 1259.,  1224.],
        [ 1256.,  1229.]]], dtype=float32)

corners

[array([[[ 1339.,   951.],
         [ 1413.,   934.],
         [ 1434.,   981.],
         [ 1358.,   999.]]], dtype=float32), array([[[ 2247.,  1604.],
         [ 2306.,  1653.],
         [ 2263.,  1691.],
         [ 2203.,  1643.]]], dtype=float32), array([[[ 2071.,  1279.],
         [ 2101.,  1233.],
         [ 2162.,  1267.],
         [ 2132.,  1314.]]], dtype=float32), array([[[ 1209.,  1217.],
         [ 1297.,  1218.],
         [ 1290.,  1287.],
         [ 1201.,  1286.]]], dtype=float32), array([[[ 1507.,  1244.],
         [ 1510.,  1309.],
         [ 1421.,  1313.],
         [ 1419.,  1245.]]], dtype=float32), array([[[  940.,  1212.],
         [  933.,  1282.],
         [  840.,  1285.],
         [  849.,  1216.]]], dtype=float32), array([[[ 2736.,  1132.],
         [ 2764.,  1183.],
         [ 2723.,  1241.],
         [ 2701.,  1191.]]], dtype=float32), array([[[ 1140.,  1120.],
         [ 1129.,  1059.],
         [ 1214.,  1048.],
         [ 1226.,  1108.]]], dtype=float32), array([[[  990.,  1050.],
         [  906.,  1071.],
         [  885.,  1013.],
         [  968.,   993.]]], dtype=float32), array([[[ 1586.,   950.],
         [ 1513.,   929.],
         [ 1543.,   879.],
         [ 1616.,   899.]]], dtype=float32)]

Pretty fast processing !

4: Results¶

plt.figure()
plt.imshow(frame_markers, origin = "upper")
if ids is not None:
    for i in range(len(ids)):
        c = corners[i][0]
        plt.plot([c[:, 0].mean()], [c[:, 1].mean()], "+", label = "id={0}".format(ids[i]))
"""for points in rejectedImgPoints:
    y = points[:, 0]
    x = points[:, 1]
    plt.plot(x, y, ".m-", linewidth = 1.)"""
plt.legend()
plt.show()

<IPython.core.display.Javascript object>

def quad_area(data):
    l = data.shape[0]//2
    corners = data[["c1", "c2", "c3", "c4"]].values.reshape(l, 2,4)
    c1 = corners[:, :, 0]
    c2 = corners[:, :, 1]
    c3 = corners[:, :, 2]
    c4 = corners[:, :, 3]
    e1 = c2-c1
    e2 = c3-c2
    e3 = c4-c3
    e4 = c1-c4
    a = -.5 * (np.cross(-e1, e2, axis = 1) + np.cross(-e3, e4, axis = 1))
    return a

corners2 = np.array([c[0] for c in corners])

data = pd.DataFrame({"x": corners2[:,:,0].flatten(), "y": corners2[:,:,1].flatten()},
                   index = pd.MultiIndex.from_product(
                           [ids.flatten(), ["c{0}".format(i )for i in np.arange(4)+1]],
                       names = ["marker", ""] ))

data = data.unstack().swaplevel(0, 1, axis = 1).stack()
data["m1"] = data[["c1", "c2"]].mean(axis = 1)
data["m2"] = data[["c2", "c3"]].mean(axis = 1)
data["m3"] = data[["c3", "c4"]].mean(axis = 1)
data["m4"] = data[["c4", "c1"]].mean(axis = 1)
data["o"] = data[["m1", "m2", "m3", "m4"]].mean(axis = 1)
data

		c1	c2	c3	c4	m1	m2	m3	m4	o
marker
1	x	1209.0	1297.0	1290.0	1201.0	1253.0	1293.5	1245.5	1205.0	1249.25
1	y	1217.0	1218.0	1287.0	1286.0	1217.5	1252.5	1286.5	1251.5	1252.00
3	x	2736.0	2764.0	2723.0	2701.0	2750.0	2743.5	2712.0	2718.5	2731.00
3	y	1132.0	1183.0	1241.0	1191.0	1157.5	1212.0	1216.0	1161.5	1186.75
4	x	1140.0	1129.0	1214.0	1226.0	1134.5	1171.5	1220.0	1183.0	1177.25
4	y	1120.0	1059.0	1048.0	1108.0	1089.5	1053.5	1078.0	1114.0	1083.75
5	x	2071.0	2101.0	2162.0	2132.0	2086.0	2131.5	2147.0	2101.5	2116.50
5	y	1279.0	1233.0	1267.0	1314.0	1256.0	1250.0	1290.5	1296.5	1273.25
6	x	1507.0	1510.0	1421.0	1419.0	1508.5	1465.5	1420.0	1463.0	1464.25
6	y	1244.0	1309.0	1313.0	1245.0	1276.5	1311.0	1279.0	1244.5	1277.75
7	x	2247.0	2306.0	2263.0	2203.0	2276.5	2284.5	2233.0	2225.0	2254.75
7	y	1604.0	1653.0	1691.0	1643.0	1628.5	1672.0	1667.0	1623.5	1647.75
9	x	940.0	933.0	840.0	849.0	936.5	886.5	844.5	894.5	890.50
9	y	1212.0	1282.0	1285.0	1216.0	1247.0	1283.5	1250.5	1214.0	1248.75
10	x	990.0	906.0	885.0	968.0	948.0	895.5	926.5	979.0	937.25
10	y	1050.0	1071.0	1013.0	993.0	1060.5	1042.0	1003.0	1021.5	1031.75
11	x	1339.0	1413.0	1434.0	1358.0	1376.0	1423.5	1396.0	1348.5	1386.00
11	y	951.0	934.0	981.0	999.0	942.5	957.5	990.0	975.0	966.25
12	x	1586.0	1513.0	1543.0	1616.0	1549.5	1528.0	1579.5	1601.0	1564.50
12	y	950.0	929.0	879.0	899.0	939.5	904.0	889.0	924.5	914.25

# Plante un peu...
"""cap = cv2.VideoCapture('_data/AeroTrain.mp4')
while(cap.isOpened()):
    ret, frame = cap.read()

    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    cv2.imshow('frame',gray)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()"""

"cap = cv2.VideoCapture('_data/AeroTrain.mp4')nwhile(cap.isOpened()):n    ret, frame = cap.read()nn    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)nn    cv2.imshow('frame',gray)n    if cv2.waitKey(1) & 0xFF == ord('q'):n        breaknncap.release()ncv2.destroyAllWindows()"

cap = cv2.VideoCapture('_data/AeroTrain.mp4')
nframe = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))

print("nframe =", nframe)
cap.set(1, 300) # arguments: 1: laisser, 2: numéro du frame
ret, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
plt.figure()
plt.imshow(gray)
plt.show()
cap.release()

nframe = 712

<IPython.core.display.Javascript object>

%%time
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
aruco_dict = aruco.Dictionary_get(aruco.DICT_6X6_250)
parameters =  aruco.DetectorParameters_create()
corners, ids, rejectedImgPoints = aruco.detectMarkers(gray, aruco_dict, parameters=parameters)
frame_markers = aruco.drawDetectedMarkers(frame.copy(), corners, ids)

Wall time: 31.3 ms

plt.figure()
plt.imshow(frame_markers, origin = "upper")
if ids is not None:
    for i in range(len(ids)):
        c = corners[i][0]
        plt.plot([c[:, 0].mean()], [c[:, 1].mean()], "+", label = "id={0}".format(ids[i]))
"""for points in rejectedImgPoints:
    y = points[:, 0]
    x = points[:, 1]
    plt.plot(x, y, ".m-", linewidth = 1.)"""
plt.legend()
plt.show()

<IPython.core.display.Javascript object>

help(aruco.DetectorParameters_create)

Help on built-in function DetectorParameters_create:

DetectorParameters_create(...)
    DetectorParameters_create() -> retval
    .