1
Pose Estimation {#tutorial_py_pose}
8
- We will learn to exploit calib3d module to create some 3D effects in images.
13
This is going to be a small section. During the last session on camera calibration, you have found
14
the camera matrix, distortion coefficients etc. Given a pattern image, we can utilize the above
15
information to calculate its pose, or how the object is situated in space, like how it is rotated,
16
how it is displaced etc. For a planar object, we can assume Z=0, such that, the problem now becomes
17
how camera is placed in space to see our pattern image. So, if we know how the object lies in the
18
space, we can draw some 2D diagrams in it to simulate the 3D effect. Let's see how to do it.
20
Our problem is, we want to draw our 3D coordinate axis (X, Y, Z axes) on our chessboard's first
21
corner. X axis in blue color, Y axis in green color and Z axis in red color. So in-effect, Z axis
22
should feel like it is perpendicular to our chessboard plane.
24
First, let's load the camera matrix and distortion coefficients from the previous calibration
31
# Load previously saved data
32
with np.load('B.npz') as X:
33
mtx, dist, _, _ = [X[i] for i in ('mtx','dist','rvecs','tvecs')]
35
Now let's create a function, draw which takes the corners in the chessboard (obtained using
36
**cv2.findChessboardCorners()**) and **axis points** to draw a 3D axis.
38
def draw(img, corners, imgpts):
39
corner = tuple(corners[0].ravel())
40
img = cv2.line(img, corner, tuple(imgpts[0].ravel()), (255,0,0), 5)
41
img = cv2.line(img, corner, tuple(imgpts[1].ravel()), (0,255,0), 5)
42
img = cv2.line(img, corner, tuple(imgpts[2].ravel()), (0,0,255), 5)
45
Then as in previous case, we create termination criteria, object points (3D points of corners in
46
chessboard) and axis points. Axis points are points in 3D space for drawing the axis. We draw axis
47
of length 3 (units will be in terms of chess square size since we calibrated based on that size). So
48
our X axis is drawn from (0,0,0) to (3,0,0), so for Y axis. For Z axis, it is drawn from (0,0,0) to
49
(0,0,-3). Negative denotes it is drawn towards the camera.
51
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 30, 0.001)
52
objp = np.zeros((6*7,3), np.float32)
53
objp[:,:2] = np.mgrid[0:7,0:6].T.reshape(-1,2)
55
axis = np.float32([[3,0,0], [0,3,0], [0,0,-3]]).reshape(-1,3)
57
Now, as usual, we load each image. Search for 7x6 grid. If found, we refine it with subcorner
58
pixels. Then to calculate the rotation and translation, we use the function,
59
**cv2.solvePnPRansac()**. Once we those transformation matrices, we use them to project our **axis
60
points** to the image plane. In simple words, we find the points on image plane corresponding to
61
each of (3,0,0),(0,3,0),(0,0,3) in 3D space. Once we get them, we draw lines from the first corner
62
to each of these points using our draw() function. Done !!!
64
for fname in glob.glob('left*.jpg'):
65
img = cv2.imread(fname)
66
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
67
ret, corners = cv2.findChessboardCorners(gray, (7,6),None)
70
corners2 = cv2.cornerSubPix(gray,corners,(11,11),(-1,-1),criteria)
72
# Find the rotation and translation vectors.
73
rvecs, tvecs, inliers = cv2.solvePnPRansac(objp, corners2, mtx, dist)
75
# project 3D points to image plane
76
imgpts, jac = cv2.projectPoints(axis, rvecs, tvecs, mtx, dist)
78
img = draw(img,corners2,imgpts)
80
k = cv2.waitKey(0) & 0xff
82
cv2.imwrite(fname[:6]+'.png', img)
84
cv2.destroyAllWindows()
86
See some results below. Notice that each axis is 3 squares long.:
88
![image](images/pose_1.jpg)
92
If you want to draw a cube, modify the draw() function and axis points as follows.
94
Modified draw() function:
96
def draw(img, corners, imgpts):
97
imgpts = np.int32(imgpts).reshape(-1,2)
99
# draw ground floor in green
100
img = cv2.drawContours(img, [imgpts[:4]],-1,(0,255,0),-3)
102
# draw pillars in blue color
103
for i,j in zip(range(4),range(4,8)):
104
img = cv2.line(img, tuple(imgpts[i]), tuple(imgpts[j]),(255),3)
106
# draw top layer in red color
107
img = cv2.drawContours(img, [imgpts[4:]],-1,(0,0,255),3)
111
Modified axis points. They are the 8 corners of a cube in 3D space:
113
axis = np.float32([[0,0,0], [0,3,0], [3,3,0], [3,0,0],
114
[0,0,-3],[0,3,-3],[3,3,-3],[3,0,-3] ])
116
And look at the result below:
118
![image](images/pose_2.jpg)
120
If you are interested in graphics, augmented reality etc, you can use OpenGL to render more