Ask Your Question

Adjust image rotation to camera rotation

asked 2018-08-07 21:14:28 -0500

Raki gravatar image

updated 2018-08-07 21:31:40 -0500


I have written a program which blends a space shuttle icon into an image which is projected. My program, additionally, uses feature matching, homography and perspective transform so that the icon can be moved as the camera moves. In other words, move the camera over the projected image to move the icon around.

I am done with everything describe above. One thing I want to accomplish is to rotate the icon image with the camera. That is, do not always draw the icon as it is, rotate it to left and right when the camera rotates so.

Here is my code which blends the icon into the image:

# blend the shuttleIcon into the image
        processedImage = cv2.cvtColor(processedImage, cv2.COLOR_BGR2BGRA)
        w, h, c = shuttleIcon.shape

        for i in range(0, w):
            for j in range(0, h):
                if shuttleIcon[i, j][3] != 0:
                    processedImage[trackedCenterPoint[1] + i, trackedCenterPoint[0] + j] = shuttleIcon[i, j]

        processedImage = cv2.cvtColor(processedImage, cv2.COLOR_BGRA2BGR)

How can I incorporate the camera rotation into this? Any thoughts?

NOTE: I do not have the calibration matrix.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2018-08-08 03:04:43 -0500

Raki gravatar image

OK. I got it. The following does the job:

'''Takes 2 vectors and returns the rotation matrix between these 2 vectors'''
def get_camera_rotation(homographyMatrix):
    # Points in the camera frame
    camera_pts = np.float32([[round(CAM_WIDTH / 2), round(CAM_HEIGHT / 2)],
                             [round(10 + CAM_WIDTH / 2), round(10 + CAM_HEIGHT / 2)]]).reshape(-1, 1, 2)
    # Find these points in the projector image
    proj_pts = cv2.perspectiveTransform(camera_pts, homographyMatrix)

    # Find the vectors between the sets of points
    camera_vector = (camera_pts[0][0][0] - camera_pts[1][0][0], camera_pts[0][0][1] - camera_pts[1][0][1])
    proj_vector = (proj_pts[0][0][0] - proj_pts[1][0][0], proj_pts[0][0][1] - proj_pts[1][0][1])

    # change the vectors to unit vectors
    camera_vector = camera_vector / np.absolute(np.linalg.norm(camera_vector))
    proj_vector = proj_vector / np.absolute(np.linalg.norm(proj_vector))

    # calculate the angle between the 2 vectors
    # Change the sign of the angle if the rocket is turning the opposite way to desired
    #sine of the angle
    sinAngle = camera_vector[0] * proj_vector[1] - camera_vector[1] * proj_vector[0]
    #angle between the vectors
    angle = np.arcsin(np.clip(sinAngle, -1.0, 1.0))

    #print("The angle between the camera and the projector is:")
    # calculate the 2D rotation matrix from this angle
    rotation_matrix = np.matrix([[np.cos(angle), -1 * np.sin(angle)], [np.sin(angle), np.cos(angle)]])
    return angle

to use:

 # rotate the shuttle as the camera does
        # first though, get a copy
        toBeRotatedShuttle = shuttleIcon.copy()
        rows, cols, w = toBeRotatedShuttle.shape
        angle = get_camera_rotation(smoothenedMatrix)
        angleInDegrees = round(math.degrees(math.asin(angle)),2) # convert radian to degrees
        rotationMatrix = cv2.getRotationMatrix2D((cols / 2, rows / 2), angleInDegrees, 1)
        shuttleIcon = cv2.warpAffine(toBeRotatedShuttle, rotationMatrix, (cols, rows), cv2.INTER_AREA)
edit flag offensive delete link more
Login/Signup to Answer

Question Tools

1 follower


Asked: 2018-08-07 21:14:28 -0500

Seen: 401 times

Last updated: Aug 08 '18