-
Notifications
You must be signed in to change notification settings - Fork 4
/
vmarker.py
165 lines (145 loc) · 6.3 KB
/
vmarker.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
#!/usr/bin/env python
# -*- coding: utf-8 -*
# assuming python2 for ROS
import cv2
import numpy as np
import math
def create_marker(mnum,imsize=200):
aruco = cv2.aruco
dictionary = aruco.getPredefinedDictionary(aruco.DICT_6X6_250)
markers = []
for i in range(mnum):
marker = aruco.drawMarker(dictionary, i, imsize)
markers.append(marker)
cv2.imwrite('markers/marker'+str(i)+'.png', marker)
def detect_marker(frame):
aruco = cv2.aruco
dictionary = aruco.getPredefinedDictionary(aruco.DICT_6X6_250)
corners, ids, rejectedImgPoints = aruco.detectMarkers(frame, dictionary)
dimage = aruco.drawDetectedMarkers(frame,corners)
return dimage
class vmarker:
def __init__(self,markernum=5,K=[],dist=[],markerpos_file="mpos1.csv"):
aruco = cv2.aruco
self.dictionary = aruco.getPredefinedDictionary(aruco.DICT_6X6_250)
#self.startvideo()
self.mnum = markernum
self.setmarker(markerpos_file)
self.K = K
self.dist = dist
self.tvecs = []
self.rvecs = []
self.R = []
self.PNPsolved = False
self.hasCameraPose = False
def setmarker(self,fname):
#self.objp = np.zeros((markernum,3), np.float32)
self.objp = np.loadtxt(fname,delimiter=",")
self.mnum , _ = self.objp.shape
print(self.mnum)
def startvideo(self,vnum=0):
self.cap = cv2.VideoCapture(0)
def showmarker(self,frame):
aruco = cv2.aruco
corners, ids, rejectedImgPoints = aruco.detectMarkers(frame, self.dictionary)
detect = aruco.drawDetectedMarkers(frame,corners)
cv2.imshow("detected",detect)
cv2.waitKey(1)
def load_camerapose_yml(self,file):
try:
import yaml
with open(file, 'r+') as stream:
dic = yaml.load(stream)
print(dic)
self.rvecs = np.float32(dic["rvecs"]).reshape(3,1)
self.tvecs = np.float32(dic["tvecs"]).reshape(3,1)
self.PNPsolved = True
except:
print("Something going wrong!")
def getcamerapose(self,frame):
aruco = cv2.aruco
corners, ids, rejectedImgPoints = aruco.detectMarkers(frame, self.dictionary)
self.PNPsolved = False
if len(corners) == self.mnum:
# sort based on IDs and use center value
centercorners = []
geometrypositions = []
for id_,corner in sorted(zip(ids,corners)): #corner=[x11,y11]...
centercorners.append(np.average(corner,1))
geometrypositions.append(self.objp[id_])
self.ccorners = np.array(centercorners).reshape(self.mnum,1,2)
self.realcornerpos = np.array(geometrypositions).reshape(len(ids),3)
#print(self.ccorners)
# Find the rotation and translation vectors.
self.PNPsolved, self.rvecs, self.tvecs, inliers = cv2.solvePnPRansac(self.realcornerpos, self.ccorners, self.K, self.dist)
self.hasCameraPose = True
self.drawaxis(aruco.drawDetectedMarkers(frame,corners,ids)) # draw origin
self.R,_ = cv2.Rodrigues(self.rvecs)
return -np.dot(self.R.T,self.tvecs)
else:
if self.hasCameraPose:
self.drawaxis(frame)
return -np.dot(self.R.T,self.tvecs)
else:
cv2.imshow('projected',aruco.drawDetectedMarkers(frame,corners,ids))
cv2.waitKey(1)
return []
def draw(self,img, origin, imgpts):
corner = tuple(origin[0].ravel())
img = cv2.line(img, corner, tuple(imgpts[0].ravel()), (255,0,0), 5)
img = cv2.line(img, corner, tuple(imgpts[1].ravel()), (0,255,0), 5)
img = cv2.line(img, corner, tuple(imgpts[2].ravel()), (0,0,255), 5)
return img
def drawaxis(self,frame):#30cm cube
self.axis = np.float32([[0.3,0,0], [0,0.3,0], [0,0,0.3]]).reshape(-1,3)
self.origin = np.float32([[0,0,0]]).reshape(-1,3)
imgpts, jac = cv2.projectPoints(self.axis, self.rvecs, self.tvecs, self.K, self.dist)
imgorgs, _ = cv2.projectPoints(self.origin, self.rvecs, self.tvecs, self.K, self.dist)
img = self.draw(frame,imgorgs,imgpts)
cv2.imshow('projected',img)
cv2.waitKey(1)
# z: object height
def getobjpose_1(self,objpts,z):
self.R,_ = cv2.Rodrigues(self.rvecs)
pt = cv2.undistortPoints(np.array(objpts).reshape(-1,1,2),self.K,self.dist,P=self.K)
self.Rt = np.concatenate([self.R,self.tvecs],axis=1)
# Extraction
self.P = np.dot(self.K,self.Rt)
A3 = - np.float32([pt[0,0,0],pt[0,0,1],1]).reshape(3,1) #A1,A2 = self.Rt[:,0],self.Rt[:,1]
A4 = self.P[:,2:3]*z+self.P[:,3:4]
A = np.concatenate([self.P[:,0:2],A3,A4],axis=1)
try:
U, S, V = np.linalg.svd(np.dot(A.T,A)) # use svd to get null space
except: # see here http://oppython.hatenablog.com/entry/2014/01/21/003245
S2,vt = np.linalg.eigh(np.dot(A.T ,A)) # if numpy method is unstable
vt=vt[:,::-1]#,S2 = w[::-1]
V = vt.T
vec = V[3]
X = vec[0]/ vec[3]
Y = vec[1]/ vec[3]
return [X,Y]
# assume zero height
def getobjpose_2(self,objpts):
# get 3d pose and convert to 2d
plane2dmap = self.objp[:,0:2].reshape(-1,1,2)
# then extract homography from 2dpose and image points
Homo,inliner = cv2.findHomography(self.ccorners,plane2dmap,cv2.RANSAC,3.0)
# finally from using homography to convert observed pts to 3d pose
pos=cv2.perspectiveTransform(np.float32([cx,cy]).reshape(-1,1,2),Homo)
return pos[0,0]
if __name__=='__main__':
cap = cv2.VideoCapture(0)
# load camera matrix and distort matrix
K = np.loadtxt("calib_usb/K.csv",delimiter=",")
dist_coef = np.loadtxt('calib_usb/d.csv',delimiter=",")
vm = vmarker(K=K,dist=dist_coef,markerpos_file="sample/markers1to4.csv")
try:
while ~cap.isOpened():
ok,frame = cap.read()
nframe = cv2.undistort(frame, K, dist_coef)
tv = vm.getcamerapose(frame)
print(tv)
#print(vm.rvec) #euler angle
except KeyboardInterrupt:
print("Finish Program!")
exit(0)