def get_device_fy(config):
from realsense_device_manager import DeviceManager
device_manager = DeviceManager(rs.context(), config)
device_manager.enable_all_devices()
# Allow some frames for the auto-exposure controller to stablise
for frame in range(30):
frames = device_manager.poll_frames()
assert (len(device_manager._available_devices) > 0)
intrinsics_devices = device_manager.get_device_intrinsics(frames)
temp = intrinsics_devices['908212070032']
temp2 = temp[rs.stream.color].__getattribute__('fy')
device_manager.disable_streams()
return temp2
def run():
rs_config = rs.config()
rs_config.enable_stream(rs.stream.color, CAMERA_RESOLUTION_WIDTH,
CAMERA_RESOLUTION_HEIGHT, rs.format.bgr8,
CAMERA_FRAME_RATE)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
assert (len(device_manager._available_devices) > 0)
try:
while True:
frames = device_manager.poll_frames()
color_images = []
for key in frames.keys():
color_frame = frames[key][rs.stream.color].get_data()
color_image = np.asanyarray(color_frame)
color_images.append(color_image)
if len(color_images) == 0:
continue
images = np.vstack(color_images)
cv2.namedWindow(CV_WINDOW_NAME, cv2.WINDOW_NORMAL)
cv2.resizeWindow(CV_WINDOW_NAME, INIT_WINDOW_WIDTH,
INIT_WINDOW_HEIGHT)
cv2.imshow(CV_WINDOW_NAME, images)
k = cv2.waitKey(1)
if k == 112: # 'p' key
print("Capturing images...")
for color_image in color_images:
status, dir = captureFrame(color_image)
if status: print("Saved image to " + dir)
else: print("A problem occurred when capturing the frame.")
elif k == 27:
break
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
def main():
# First we set up the cameras to start streaming
# Define some constants
resolution_width = 1280 # pixels
resolution_height = 720 # pixels
frame_rate = 30 # fps
dispose_frames_for_stablisation = 30 # frames
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width,
resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width,
resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width,
resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
device_manager._available_devices.sort()
# Allow some frames for the auto-exposure controller to stablise
for frame in range(dispose_frames_for_stablisation):
frames = device_manager.poll_frames()
assert (len(device_manager._available_devices) > 0)
#Then we calibrate the images
# Get the intrinsics of the realsense device
intrinsics_devices = device_manager.get_device_intrinsics(frames)
# Set the charuco board parameters for calibration
aruco_dict = aruco.Dictionary_get(aruco.DICT_6X6_250)
charuco_width = 8
charuco_height = 5
square_length = 0.03425
marker_length = .026
coordinate_dimentions = 3
charuco_board = aruco.CharucoBoard_create(charuco_width, charuco_height,
square_length, marker_length,
aruco_dict)
# Estimate the pose of the cameras compared to the first camera in the list
amount_devices = len(device_manager._available_devices)
transformation_matrix = {}
rms_matrix = {}
for device in device_manager._available_devices:
transformation_matrix[device] = {}
rms_matrix[device] = {}
for device2 in device_manager._available_devices:
rms_matrix[device][device2] = np.inf
devices_stitched = False
while not devices_stitched:
frames = device_manager.poll_frames()
pose_estimator = PoseEstimation(frames, intrinsics_devices,
charuco_board)
transformation_result_kabsch = pose_estimator.perform_pose_estimation()
object_point = pose_estimator.get_chessboard_corners_in3d()
calibrated_device_count = 0
for device in device_manager._available_devices:
if not transformation_result_kabsch[device][0]:
print("Device", device, "needs to be calibrated")
else:
# If this is the first camera in the list
if calibrated_device_count == 0:
source_matrix = transformation_result_kabsch[device][1]
source_device = device
source_rms = transformation_result_kabsch[device][3]
else:
# If new estimate is better than previous
if source_rms + transformation_result_kabsch[device][
3] < rms_matrix[device][source_device]:
rms_matrix[source_device][
device] = source_rms + transformation_result_kabsch[
device][3]
rms_matrix[device][
source_device] = source_rms + transformation_result_kabsch[
device][3]
slave_transfrom = transformation_result_kabsch[device][
1].inverse()
multiplied_transform = np.matmul(
source_matrix.pose_mat, slave_transfrom.pose_mat)
Multiplied_transform = Transformation(
multiplied_transform[:3, :3],
multiplied_transform[:3, 3])
transformation_matrix[device][
source_device] = Multiplied_transform
temp_inverted_matrix = np.matmul(
source_matrix.pose_mat, slave_transfrom.pose_mat)
inverted_transform = Transformation(
temp_inverted_matrix[:3, :3],
temp_inverted_matrix[:3, 3])
transformation_matrix[source_device][
device] = inverted_transform.inverse()
calibrated_device_count += 1
# Check if all devices are stitched together
transformation_devices = least_error_transfroms(
transformation_matrix, rms_matrix)
if transformation_devices != 0:
devices_stitched = True
test = matrix_viewer(rms_matrix)
print(test)
print("Calibration completed... \n")
# Enable the emitter of the devices and extract serial numbers to identify cameras
device_manager.enable_emitter(True)
key_list = device_manager.poll_frames().keys()
pcd = o3d.geometry.PointCloud()
# enable visualiser
vis = o3d.visualization.Visualizer()
vis.create_window()
first_image = True
#Stitch together all the different camera pointclouds from different cameras
pcs = {}
for camera in key_list:
pcs[camera] = rs.pointcloud()
pixels = resolution_width * resolution_height
total_pixels = pixels * len(key_list)
cloud = np.zeros((3, total_pixels))
transformed_pixels = np.ones((4, pixels))
idxe = np.random.permutation(cloud.shape[1])
while True:
start = time.time()
the_frames = device_manager.poll_frames()
for idx, camera in enumerate(key_list):
pc = pcs[camera]
frame = the_frames[camera][rs.stream.depth]
points = pc.calculate(frame)
vert = np.transpose(np.asanyarray(points.get_vertices(2)))
transformed_pixels[:coordinate_dimentions, :] = vert
calibrated_points = np.matmul(transformation_devices[camera],
transformed_pixels)
cloud[:, pixels * idx:pixels *
(idx + 1)] = calibrated_points[:coordinate_dimentions, :]
# Reduces rendered points and removes points with extreme z values
keep_ratio = 0.01
cloud_filtered = cloud[:,
idxe[0:math.floor(cloud.shape[1] * keep_ratio)]]
#cloud_filtered = cloud_filtered - np.min(cloud_filtered[2, :])
dist_thresh = 3
cloud_filtered = -cloud_filtered[:, cloud_filtered[2, :] < dist_thresh]
# cloud_filtered = cloud_filtered[:, cloud_filtered[2, :] > -1]
# cloud_filtered = cloud_filtered[:, np.invert(np.any(cloud_filtered > dist_thresh, axis=0))]
# cloud_filtered = cloud_filtered[:, np.invert(np.any(cloud_filtered > dist_thresh, axis=0))]
# renders points from all different cameras
#mlab.points3d( cloud_filtered[0, :], cloud_filtered[1, :], cloud_filtered[2, :], scale_factor=0.1)
#mlab.show()
pcd.points = o3d.utility.Vector3dVector(np.transpose(cloud_filtered))
if first_image:
vis.add_geometry(pcd)
first_image = False
vis.update_geometry()
vis.poll_events()
vis.update_renderer()
end = time.time()
#txt = input()
#print(1 / (end - start))
device_manager.disable_streams()
vis.destroy_window()
# Define some constants
resolution_width = 640 # pixels
resolution_height = 480 # pixels
frame_rate = 15 # fps
dispose_frames_for_stablisation = 30 # frames
try:
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width, resolution_height, rs.format.z16, frame_rate)
# rs_config.enable_stream(rs.stream.infrared, 1, resolution_width, resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width, resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
# Allow some frames for the auto-exposure controller to stablise
while True:
frames = device_manager.poll_frames()
visualise_measurements_cv2(frames)
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
def run_demo():
# Define some constants
resolution_width = 1280 # pixels
resolution_height = 720 # pixels
frame_rate = 15 # fps
dispose_frames_for_stablisation = 30 # frames
# chessboard_width = 6 # squares
# chessboard_height = 9 # squares
# square_size = 0.0253 # meters
try:
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width, resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width, resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width, resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
# Allow some frames for the auto-exposure controller to stablise
for frame in range(dispose_frames_for_stablisation):
frames = device_manager.poll_frames()
assert( len(device_manager._available_devices) > 0 )
"""
1: Calibration
Calibrate all the available devices to the world co-ordinates.
For this purpose, a chessboard printout for use with opencv based calibration process is needed.
"""
# Get the intrinsics of the realsense device
intrinsics_devices = device_manager.get_device_intrinsics(frames)
# Set the chessboard parameters for calibration
# chessboard_params = [chessboard_height, chessboard_width, square_size]
# Estimate the pose of the chessboard in the world coordinate using the Kabsch Method
# calibrated_device_count = 0
# while calibrated_device_count < len(device_manager._available_devices):
# frames = device_manager.poll_frames()
# pose_estimator = PoseEstimation(frames, intrinsics_devices, chessboard_params)
# transformation_result_kabsch = pose_estimator.perform_pose_estimation()
# object_point = pose_estimator.get_chessboard_corners_in3d()
# calibrated_device_count = 0
# for device in device_manager._available_devices:
# if not transformation_result_kabsch[device][0]:
# print("Place the chessboard on the plane where the object needs to be detected..")
# else:
# calibrated_device_count += 1
# Save the transformation object for all devices in an array to use for measurements
# transformation_devices={}
# chessboard_points_cumulative_3d = np.array([-1,-1,-1]).transpose()
# for device in device_manager._available_devices:
# transformation_devices[device] = transformation_result_kabsch[device][1].inverse()
# points3D = object_point[device][2][:,object_point[device][3]]
# points3D = transformation_devices[device].apply_transformation(points3D)
# chessboard_points_cumulative_3d = np.column_stack( (chessboard_points_cumulative_3d,points3D) )
# # Extract the bounds between which the object's dimensions are needed
# # It is necessary for this demo that the object's length and breath is smaller than that of the chessboard
# chessboard_points_cumulative_3d = np.delete(chessboard_points_cumulative_3d, 0, 1)
# roi_2D = get_boundary_corners_2D(chessboard_points_cumulative_3d)
print("Calibration completed... \nPlace the box in the field of view of the devices...")
"""
2: Measurement and display
Measure the dimension of the object using depth maps from multiple RealSense devices
The information from Phase 1 will be used here
"""
# Enable the emitter of the devices
device_manager.enable_emitter(True)
# Load the JSON settings file in order to enable High Accuracy preset for the realsense
device_manager.load_settings_json("./HighResHighAccuracyPreset.json")
# Get the extrinsics of the device to be used later
extrinsics_devices = device_manager.get_depth_to_color_extrinsics(frames)
# Get the calibration info as a dictionary to help with display of the measurements onto the color image instead of infra red image
calibration_info_devices = defaultdict(list)
for calibration_info in (transformation_devices, intrinsics_devices, extrinsics_devices):
for key, value in calibration_info.items():
calibration_info_devices[key].append(value)
# Continue acquisition until terminated with Ctrl+C by the user
while 1:
# Get the frames from all the devices
frames_devices = device_manager.poll_frames()
# Calculate the pointcloud using the depth frames from all the devices
point_cloud = calculate_cumulative_pointcloud(frames_devices, calibration_info_devices, roi_2D)
# Get the bounding box for the pointcloud in image coordinates of the color imager
bounding_box_points_color_image, length, width, height = calculate_boundingbox_points(point_cloud, calibration_info_devices )
# Draw the bounding box points on the color image and visualise the results
visualise_measurements(frames_devices, bounding_box_points_color_image, length, width, height)
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
def run():
try:
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width,
resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width,
resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width,
resolution_height, rs.format.bgr8, frame_rate)
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
# device_manager.enable_device(u'819612070850',False)
# pdb.set_trace()
print 'Cam Init...'
for frame in range(dispose_frames_for_stablisation):
frames = device_manager.poll_frames()
intrinsics_devices = device_manager.get_device_intrinsics(frames)
# pdb.set_trace()
t0 = time.time()
q = (cv2.waitKey(1)) & 0xFF
imgs_fanuc, imgs_ur = [], []
times = []
vels = []
ur_poses = []
i = 0
t_pub = time.time()
save_img = False
while q != 27:
frames = device_manager.poll_frames()
if frames is {} or len(frames) != 2:
continue
# print device_manager._enabled_devices
if q == ord('s'):
save_img = True
start_time = time.time()
for serial in device_manager._enabled_devices:
color_img = np.array(
frames[serial][rs.stream.color].get_data())
cv2.imshow(win_serial[serial], color_img)
if serial == u'818312071299':
img_ur = color_img
time_ur = time.time()
elif serial == u'819612070850':
img_fanuc = color_img
time_fanuc = time.time()
# print "fanuc_camera is coming......"
# print '% d camera time : %f '%(len(device_manager._enabled_devices),time.time()-t0)
q = (cv2.waitKey(10)) & 0xFF
if save_img:
imgs_fanuc.append(img_fanuc)
imgs_ur.append(img_ur)
times.append(time_fanuc - start_time)
ur_poses.append(get_frame('tool0'))
t0 = time.time()
if time.time() - t_pub > 0.15:
ret, img, corner = detect_grid(img_fanuc)
if ret:
rvec, tvec, error = pose_estimation(corner)
tvec = tvec / 1000.
# print 'error:',error
Pose2_H = XYZRodrigues_to_Hmatrix(
tvec.flatten(1).tolist() + rvec.flatten(1).tolist())
# print "Detection error is %f" % error
else:
print "Detection of chess point is wrong."
continue
temp = np.dot(ur2fanuc_H, fanuc_hand_to_eye_H)
temp = np.dot(temp, Pose2_H)
temp = np.dot(temp, Track_H)
target_H = np.dot(temp, np.linalg.inv(ur5_hand_in_eye_H))
target_Q = quaternion_from_matrix(target_H)
# print "ur5 target pose: "
# print "%f %f %f %f %f %f %f" % (target_H[0, 3], target_H[1, 3], target_H[2, 3],
# target_Q[0], target_Q[1], target_Q[2],
# target_Q[3])
# cv2.destroyAllWindows()
# pdb.set_trace()
tt = time.time()
# t_matrix,_ = getFrame('base','tool0')
# current_pose = Hmatrix_to_XYZRodrigues(t_matrix)
# print current_pose
current_pose = get_frame('tool0')
# print current_pose
# pdb.set_trace()
# print " get current pose cause : %f"%(time.time()-tt)
rvec, _ = cv2.Rodrigues(target_H[:3, :3])
rvec = rvec.flatten(1)
tvec = target_H[:3, 3]
target_pose = tvec.tolist() + rvec.tolist()
diff_pose = np.array(target_pose) - np.array(current_pose)
vv_t, vv_r = 3, 0.5
while np.sum((diff_pose[:3] * vv_t > 0.5).astype(np.int8)) > 0:
vv_t -= 0.1
diff_pose_vel = (diff_pose[:3] * vv_t).tolist() + (
diff_pose[3:] * vv_r).tolist()
# pdb.set_trace()
# print "----- diff pose vel ----"
# print diff_pose_vel
# pdb.set_trace()
# pose_msg = "movel(p[%5.2f, %5.2f, %5.2f, %5.2f, %5.2f, %5.2f], %5.2f, %5.2f)\n" % (
# target_H[0, 3], target_H[1, 3], target_H[2, 3],
# rvec[0], rvec[1], rvec[2], acc, vel)
diff_vel_msg = "speedl([%5.2f, %5.2f, %5.2f, %5.2f, %5.2f, %5.2f], %5.2f,%5.2f )\n" % (
diff_pose_vel[0], diff_pose_vel[1], diff_pose_vel[2],
diff_pose_vel[3], diff_pose_vel[4], diff_pose_vel[5], 1.2,
0.5)
vels.append(diff_pose_vel)
# pdb.set_trace()
pub.publish(diff_vel_msg)
print time.time() - t_pub
i += 1
t_pub = time.time()
# pdb.set_trace()
xx, yy, zz, rr, pp, qq = [], [], [], [], [], []
for vel in vels:
x, y, z, r, p, q = vel
xx.append(x)
yy.append(y)
zz.append(z)
rr.append(r)
pp.append(p)
qq.append(q)
plt.plot(xx, 'r')
plt.plot(yy, 'b')
plt.plot(zz, 'g')
# plt.figure()
plt.plot(rr, 'c')
plt.plot(pp, 'k')
plt.plot(qq, 'w')
label = ['x', 'y', 'z', 'rx', 'ry', 'rz']
plt.ylim(-0.02, 0.02)
plt.legend(label)
plt.show()
outfile_name = datetime.datetime.now().strftime(("%y%m%d_%H%M%S"))
with h5py.File(outfile_name + '_fanuc_ur_cam.hdf5', 'w') as f:
f.create_dataset('fanuc', data=np.array(imgs_fanuc))
f.create_dataset('ur', data=np.array(imgs_ur))
f.create_dataset('time', data=np.array(times))
f.create_dataset('pose', data=np.array(ur_poses))
# pub.publish(pose_msg)
# pdb.set_trace()
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
def main():
# First we set up the cameras to start streaming
# Define some constants
resolution_width = 1280 # pixels
resolution_height = 720 # pixels
frame_rate = 30 # fps
dispose_frames_for_stablisation = 30 # frames
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width,
resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width,
resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width,
resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
device_manager._available_devices.sort()
device_list = device_manager._available_devices
# Allow some frames for the auto-exposure controller to stablise
for frame in range(dispose_frames_for_stablisation):
frames = device_manager.poll_frames()
assert (len(device_manager._available_devices) > 0)
#Then we calibrate the images
# Get the intrinsics of the realsense device
intrinsics_devices = device_manager.get_device_intrinsics(frames)
# Set the charuco board parameters for calibration
aruco_dict = aruco.Dictionary_get(aruco.DICT_6X6_250)
charuco_width = 8
charuco_height = 5
square_length = 0.03425
marker_length = .026
coordinate_dimentions = 3
charuco_board = aruco.CharucoBoard_create(charuco_width, charuco_height,
square_length, marker_length,
aruco_dict)
# Choose amount of frames to average
amount_frames = 12
frame_dict = {}
transform_dict = {}
rms_dict = {}
for from_device in device_list:
transform_dict[from_device] = {}
rms_dict[from_device] = {}
for to_device in device_list:
transform_dict[from_device][to_device] = {}
rms_dict[from_device][to_device] = np.inf
devices_stitched = False
while not devices_stitched:
print("taking new set of images")
for frame_count in range(amount_frames):
print("taking image")
print(amount_frames - frame_count, "images left")
frames = device_manager.poll_frames()
print("Next image in 1 seconds")
time.sleep(1)
frame_dict[frame_count] = frames
for idx, from_device in enumerate(device_list[:-1]):
for to_device in device_list[idx + 1:]:
if to_device != from_device:
temp_transform, temp_rms = get_transformation_matrix(
frame_dict, [from_device, to_device],
intrinsics_devices, charuco_board)
if temp_rms < rms_dict[from_device][to_device]:
rms_dict[from_device][to_device] = temp_rms
rms_dict[to_device][from_device] = temp_rms
transform_dict[from_device][to_device] = temp_transform
transform_dict[to_device][
from_device] = temp_transform.inverse()
test = matrix_viewer(rms_dict)
print(test)
devices_stitched = True
for idx, from_device in enumerate(device_list[1:]):
if rms_dict[from_device][device_list[idx]] == np.inf:
devices_stitched = False
transformation_devices = {}
identity = np.identity(4)
transformation_devices[device_list[0]] = Transformation(
identity[:3, :3], identity[:3, 3])
for idx, from_device in enumerate(device_list[1:]):
temp_transform = np.matmul(
transformation_devices[device_list[idx]].pose_mat,
transform_dict[from_device][device_list[idx]].pose_mat)
transformation_devices[from_device] = Transformation(
temp_transform[:3, :3], temp_transform[:3, 3])
# Printing
print("Calibration completed... \n")
# Enable the emitter of the devices and extract serial numbers to identify cameras
device_manager.enable_emitter(True)
key_list = device_manager.poll_frames().keys()
pcd = o3d.geometry.PointCloud()
# enable visualiser
vis = o3d.visualization.Visualizer()
vis.create_window()
first_image = True
while True:
frames = device_manager.poll_frames()
displayed_points = np.zeros((10, 3))
for camera in device_list:
added_points = get_charuco_points(frames[camera],
transformation_devices[camera],
intrinsics_devices[camera],
charuco_board)
if added_points.any():
displayed_points = np.vstack(
(displayed_points, np.transpose(added_points)))
pcd.points = o3d.utility.Vector3dVector(displayed_points)
if first_image:
vis.add_geometry(pcd)
first_image = False
vis.update_geometry()
vis.poll_events()
vis.update_renderer()
device_manager.disable_streams()
vis.destroy_window()
def run_calibration(self):
try:
while True:
if world.device_manager is None:
device_manager = DeviceManager()
else:
device_manager = world.device_manager
device_manager.enable_device()
print(">CALIBRATION STARTING")
for i in range(world.stablisation):
frames = device_manager.poll_frames()
assert (device_manager._enabled_devices is not None)
intrinsic = device_manager.get_device_intrinsics(frames)
# print(type(intrinsic),intrinsic)
calibrated = False
cv.namedWindow("CALIBRATE", cv.WINDOW_AUTOSIZE)
print(">SETTING IMAGE")
while True:
frames = device_manager.poll_frames()
img = np.asanyarray(frames[rs.stream.color].get_data())
gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
found, corners = cv.findChessboardCorners(gray, (world.chessboard_width, world.chessboard_height))
if found:
cv.drawChessboardCorners(img, (world.chessboard_width, world.chessboard_height), corners, found)
cv.imshow("CALIBRATE", img)
key = cv.waitKey(1)
if key == ord('q'):
break
while calibrated == False:
frames = device_manager.poll_frames()
pose_estimator = PoseEstimation(frames, intrinsic, world.chessboard_params)
result = pose_estimator.perform_pose_estimation()
object_point = pose_estimator.get_chessboard_corners_in3d()
if not result[0]:
print("Place the chessboard on the plane where the object needs to be detected..")
else:
calibrated = True
img = np.asanyarray(frames[rs.stream.color].get_data())
cv.imshow("CALIBRATE", img)
key = cv.waitKey(1)
if key == ord('q'):
device_manager.disable_streams()
cv.destroyAllWindows()
return
trans = {}
if world.calibrate_debug:
print("matrix is: \n", result[1])
trans = result[1]
points3d = np.array([[0.0,0.3,0,0],[0.0,0,0.3,0],[0.0,0,0,-0.1]], dtype="float32")
if world.calibrate_debug:
print("world axis is:")
print(points3d)
points3d = trans.apply_transformation(points3d)
x,y = convert_pointcloud_to_depth(points3d, intrinsic[rs.stream.depth])
if world.calibrate_debug:
print("camera axis is")
print(x,y)
print("Image axis is:")
x, y = x.astype("int32"), y.astype("int32")
if world.calibrate_debug:
print(x,'\n',y)
print(object_point[2][:, object_point[3]][:, :10])
print("Chess corners is(in camera):")
print(trans.inv.apply_transformation(
object_point[2][:, object_point[3]]).T[:10])
#plot axises
while True:
color = [(255,0,0), (0,255,0), (0,0,255)]
axises = ["x", "y", "z"]
frames = device_manager.poll_frames()
img = np.asanyarray(frames[rs.stream.color].get_data())
for i in range(3):
cv.line(img, (x[0], y[0]), (x[i+1], y[i+1]), color[i], 2)
cv.putText(img, axises[i], (x[i+1], y[i+1]), cv.FONT_HERSHEY_PLAIN, 1, (0, 0, 0),2)
cv.imshow("CALIBRATE", img)
key = cv.waitKey(1)
if key == ord('q'):
print("Calibration completed... \nPlace stuffs in the field of view of the devices...")
world.world2camera = trans
self.xyAxis = np.vstack((x,y))
return
elif key == ord('r'):
break
elif key == ord('t'):
cv.imwrite("./photos/calibrate_"+world.now, img)
finally:
device_manager.disable_streams()
cv.destroyAllWindows()
def run_demo():
# Define some constants
resolution_width = 640 # pixels
resolution_height = 480 # pixels
frame_rate = 15 # fps
dispose_frames_for_stablisation = 30 # frames
chessboard_width = 6 # squares
chessboard_height = 9 # squares
square_size = 0.0253 # meters
align_to = rs.stream.color
align = rs.align(align_to)
try:
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width, resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width, resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width, resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
# Allow some frames for the auto-exposure controller to stablise
for frame in range(dispose_frames_for_stablisation):
frames,maps = device_manager.poll_frames(align)
assert( len(device_manager._available_devices) > 0 )
"""
1. Calibrate cameras and return transformation matrix (rotation matrix + translation vectors)
"""
chessboard_params = [chessboard_height, chessboard_width, square_size]
cameras = calibrateCameras(align,device_manager,frames,chessboard_params)
"""
2. Run OpenPose on each view frame
"""
# Enable the emitter of the devices
device_manager.enable_emitter(True)
# Load the JSON settings file in order to enable High Accuracy preset for the realsense
device_manager.load_settings_json("./HighResHighAccuracyPreset.json")
# Get the extrinsics of the device to be used later
extrinsics_devices = device_manager.get_depth_to_color_extrinsics(frames)
# Get the calibration info as a dictionary to help with display of the measurements onto the color image instead of infra red image
# calibration_info_devices = defaultdict(list)
# for calibration_info in (transformation_devices, intrinsics_devices, extrinsics_devices):
# for key, value in calibration_info.items():
# calibration_info_devices[key].append(value)
depth_list = {}
color_list = {}
frame_id = 0
points = {}
# Continue acquisition until terminated with Ctrl+C by the user
switch = True
while 1:
# Get the frames from all the devices
if switch:
frames_devices, maps = device_manager.poll_frames(align,500)
# print(frames_devices)
# List collector for display
depth_color= []
color = []
devices = [i for i in maps]
devices.sort()
project_depth = []
for i in devices:
# 1. Get depth map and colorize
temp = maps[i]['depth']
depth_list.setdefault(i,[])
depth_list[i].append(np.array(temp))
depth_colormap = cv2.applyColorMap(cv2.convertScaleAbs(temp, alpha=0.03), cv2.COLORMAP_JET)
depth_color.append(depth_colormap)
project_depth.append((cameras[i]['matrix'],depth_colormap))
# 2. Run OpenPose detector on image
if FLAGS_USE_BBOX:
box = bbox[i]
else:
box = None
joints,img = predict_keypoints(maps[i]['color'],box)
# 3. Save annotated color image for display
color.append(img)
color_list.setdefault(i,[])
color_list[i].append(img)
# 4. Save keypoints for that camera viewpoint
cameras[i][frame_id] = joints
# 5. Save images to folder
if FLAGS_SAVE_IMGS:
cv2.imwrite('./images/depth_{}_{}.png'.format(i,frame_id),temp)
cv2.imwrite('./images/color_{}_{}.png'.format(i,frame_id),img)
#Triangulate 3d keypoints
points[frame_id] = find3dpoints_rt(cameras,0.2,frame_id)
if points[frame_id] != 'Invalid Frame':
depth_projected = []
for img in project_depth:
points2d = project_2d(img[0],points[frame_id])
img_draw = draw_pose(points2d,'openpose',img[1],True,points[frame_id])
depth_projected.append(img_draw)
# print(img_draw.shape)
depth_color = depth_projected
# proj_img = show_img(cameras,devices[0],frame_id,points)
frame_id += 1
images = np.vstack((np.hstack(color),np.hstack(depth_color)))
# images = proj_img
# Show images for debugging
cv2.namedWindow('RealSense', cv2.WINDOW_AUTOSIZE)
cv2.imshow('RealSense', images)
key = cv2.waitKey(1)
if key == 32:
switch = not(switch)
else:
continue
# Press esc or 'q' to close the image window
if key & 0xFF == ord('q') or key == 27:
cv2.destroyAllWindows()
break
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
if FLAGS_SAVE_MATRIX:
cam_pkl = open('cameras.pkl','wb')
pkl.dump(cameras,cam_pkl)
points_pkl = open('3dpoints.pkl','wb')
pkl.dump(points,points_pkl)
def main(args):
# set up save dir
t = dt.datetime.now()
save_dir = os.path.join(ROOT, args.output_dir,
'%d-%d_%d-%d' % (t.month, t.day, t.hour, t.minute))
os.system('mkdir -p {}'.format(save_dir))
dispose_frames_for_stablisation = 20
# initialize camera node for fixed camera
pipeline = rs.pipeline()
config = rs.config()
config.enable_stream(rs.stream.depth, 1280, 720, rs.format.z16, 30)
config.enable_stream(rs.stream.color, 1280, 720, rs.format.rgb8, 30)
device_manager = DeviceManager(rs.context(), config)
available_devices = device_manager._available_devices
# create folder for each device
device_dirs = {}
for dvc in available_devices:
device_dirs[dvc] = os.path.join(save_dir, dvc)
os.system('mkdir -p {}'.format(device_dirs[dvc]))
device_manager.enable_all_devices()
# Getting the depth sensor's depth scale (see rs-align example for explanation)
depth_sensor = device_manager._enabled_devices[available_devices[
0]].pipeline_profile.get_device().first_depth_sensor()
depth_scale = depth_sensor.get_depth_scale()
print("Depth Scale is: ", depth_scale)
align_to = rs.stream.color
align = rs.align(align_to)
if args.visualize:
fig, ax = plt.subplots(2, 1)
for frame in range(dispose_frames_for_stablisation):
device_manager.poll_frames()
i = 0
now = time.time()
# Streaming loop
try:
while True:
now = time.time()
aligned_frames = {}
for dvc in available_devices:
device_manager._enabled_devices[
dvc].pipeline_profile.get_streams()
frames = device_manager._enabled_devices[
dvc].pipeline.wait_for_frames()
# Align the depth frame to color frame
aligned_frames[dvc] = align.process(frames)
print('Captured frame %07d, took %f seconds' %
(i, time.time() - now))
now = time.time()
# Get aligned frames
for dvc in aligned_frames.keys():
depth_frame = aligned_frames[dvc].get_depth_frame(
) # aligned_depth_frame is a 640x480 depth image
color_frame = aligned_frames[dvc].get_color_frame()
depth_image = np.asanyarray(
depth_frame.get_data()) * depth_scale
color_image = np.asanyarray(color_frame.get_data())
np.save(os.path.join(device_dirs[dvc], 'color_%07d.npy' % (i)),
color_image)
np.save(os.path.join(device_dirs[dvc], 'depth_%07d.npy' % (i)),
depth_image)
print('Saved frame, took %f seconds' % (time.time() - now))
# ax[0].imshow(color_image)
# ax[1].imshow(depth_image)
# plt.show()
i += 1
finally:
device_manager.disable_streams()
def run_demo():
# Define some constants
resolution_width = 1280 # pixels
resolution_height = 720 # pixels
frame_rate = 15 # fps
dispose_frames_for_stablisation = 30 # frames
chessboard_width = 6 # squares
chessboard_height = 9 # squares
square_size = 0.0253 # meters
try:
# Enable the streams from all the intel realsense devices
rs_config = rs.config()
rs_config.enable_stream(rs.stream.depth, resolution_width, resolution_height, rs.format.z16, frame_rate)
rs_config.enable_stream(rs.stream.infrared, 1, resolution_width, resolution_height, rs.format.y8, frame_rate)
rs_config.enable_stream(rs.stream.color, resolution_width, resolution_height, rs.format.bgr8, frame_rate)
# Use the device manager class to enable the devices and get the frames
device_manager = DeviceManager(rs.context(), rs_config)
device_manager.enable_all_devices()
# Allow some frames for the auto-exposure controller to stablise
for frame in range(dispose_frames_for_stablisation):
frames = device_manager.poll_frames()
assert( len(device_manager._available_devices) > 0 )
"""
1: Calibration
Calibrate all the available devices to the world co-ordinates.
For this purpose, a chessboard printout for use with opencv based calibration process is needed.
"""
# Get the intrinsics of the realsense device
intrinsics_devices = device_manager.get_device_intrinsics(frames)
# Set the chessboard parameters for calibration
chessboard_params = [chessboard_height, chessboard_width, square_size]
# Estimate the pose of the chessboard in the world coordinate using the Kabsch Method
calibrated_device_count = 0
while calibrated_device_count < len(device_manager._available_devices):
frames = device_manager.poll_frames()
pose_estimator = PoseEstimation(frames, intrinsics_devices, chessboard_params)
transformation_result_kabsch = pose_estimator.perform_pose_estimation()
object_point = pose_estimator.get_chessboard_corners_in3d()
calibrated_device_count = 0
for device in device_manager._available_devices:
if not transformation_result_kabsch[device][0]:
print("Place the chessboard on the plane where the object needs to be detected..")
else:
calibrated_device_count += 1
# Save the transformation object for all devices in an array to use for measurements
transformation_devices={}
chessboard_points_cumulative_3d = np.array([-1,-1,-1]).transpose()
for device in device_manager._available_devices:
transformation_devices[device] = transformation_result_kabsch[device][1].inverse()
points3D = object_point[device][2][:,object_point[device][3]]
points3D = transformation_devices[device].apply_transformation(points3D)
chessboard_points_cumulative_3d = np.column_stack( (chessboard_points_cumulative_3d,points3D) )
# Extract the bounds between which the object's dimensions are needed
# It is necessary for this demo that the object's length and breath is smaller than that of the chessboard
chessboard_points_cumulative_3d = np.delete(chessboard_points_cumulative_3d, 0, 1)
roi_2D = get_boundary_corners_2D(chessboard_points_cumulative_3d)
print("Calibration completed... \nPlace the box in the field of view of the devices...")
"""
2: Measurement and display
Measure the dimension of the object using depth maps from multiple RealSense devices
The information from Phase 1 will be used here
"""
# Enable the emitter of the devices
device_manager.enable_emitter(True)
# Load the JSON settings file in order to enable High Accuracy preset for the realsense
device_manager.load_settings_json("./HighResHighAccuracyPreset.json")
# Get the extrinsics of the device to be used later
extrinsics_devices = device_manager.get_depth_to_color_extrinsics(frames)
# Get the calibration info as a dictionary to help with display of the measurements onto the color image instead of infra red image
calibration_info_devices = defaultdict(list)
for calibration_info in (transformation_devices, intrinsics_devices, extrinsics_devices):
for key, value in calibration_info.items():
calibration_info_devices[key].append(value)
# Continue acquisition until terminated with Ctrl+C by the user
while 1:
# Get the frames from all the devices
frames_devices = device_manager.poll_frames()
# Calculate the pointcloud using the depth frames from all the devices
point_cloud = calculate_cumulative_pointcloud(frames_devices, calibration_info_devices, roi_2D)
# Get the bounding box for the pointcloud in image coordinates of the color imager
bounding_box_points_color_image, length, width, height = calculate_boundingbox_points(point_cloud, calibration_info_devices )
# Draw the bounding box points on the color image and visualise the results
visualise_measurements(frames_devices, bounding_box_points_color_image, length, width, height)
except KeyboardInterrupt:
print("The program was interupted by the user. Closing the program...")
finally:
device_manager.disable_streams()
cv2.destroyAllWindows()
class RealSense:
def __init__(self):
self.WIN_NAME = 'RealSense'
self.pitch, self.yaw = math.radians(-10), math.radians(-15)
self.translation = np.array([0, 0, -1], dtype=np.float32)
self.distance = 2
self.paused = False
self.decimate = 1
self.scale = True
self.color = True
# Define some constants
self.resolution_width = 1280 # pixels
self.resolution_height = 720 # pixels
self.frame_rate = 15 # fps
self.dispose_frames_for_stablisation = 25 # frames
self.chessboard_width = 6 # squares
self.chessboard_height = 9 # squares
self.square_size = 0.0253 # meters
# Allow some frames for the auto-exposure controller to stablise
self.intrinsics_devices = None
self.chessboard_params = None
self.rs_config = rs.config()
self.rs_config.enable_stream(rs.stream.depth, self.resolution_width,
self.resolution_height, rs.format.z16,
self.frame_rate)
self.rs_config.enable_stream(rs.stream.infrared, 1,
self.resolution_width,
self.resolution_height, rs.format.y8,
self.frame_rate)
self.rs_config.enable_stream(rs.stream.color, self.resolution_width,
self.resolution_height, rs.format.bgr8,
self.frame_rate)
# Use the device manager class to enable the devices and get the frames
self.device_manager = DeviceManager(rs.context(), self.rs_config)
self.device_manager.enable_all_devices()
print('0')
for self.frame in range(self.dispose_frames_for_stablisation):
self.frames = self.device_manager.poll_frames()
#print('framm = ',self.frame)
# assert( len(self.device_manager._available_devices) > 0 )
print('realsense initialized!!')
def calibaration(self):
self.intrinsics_devices = self.device_manager.get_device_intrinsics(
self.frames)
#print(' Set the chessboard parameters for calibration ')
self.chessboard_params = [
self.chessboard_height, self.chessboard_width, self.square_size
]
# Estimate the pose of the chessboard in the world coordinate using the Kabsch Method
calibrated_device_count = 0
while calibrated_device_count < len(
self.device_manager._available_devices):
self.frames = self.device_manager.poll_frames()
pose_estimator = PoseEstimation(self.frames,
self.intrinsics_devices,
self.chessboard_params)
transformation_result_kabsch = pose_estimator.perform_pose_estimation(
)
object_point = pose_estimator.get_chessboard_corners_in3d()
calibrated_device_count = 0
for device in self.device_manager._available_devices:
if not transformation_result_kabsch[device][0]:
print(
"Place the chessboard on the plane where the object needs to be detected.."
)
else:
calibrated_device_count += 1
print('calibrated_device_count =', calibrated_device_count)
# Save the transformation object for all devices in an array to use for measurements
self.transformation_devices = {}
chessboard_points_cumulative_3d = np.array([-1, -1, -1]).transpose()
for device in self.device_manager._available_devices:
self.transformation_devices[device] = transformation_result_kabsch[
device][1].inverse()
points3D = object_point[device][2][:, object_point[device][3]]
points3D = self.transformation_devices[
device].apply_transformation(points3D)
chessboard_points_cumulative_3d = np.column_stack(
(chessboard_points_cumulative_3d, points3D))
print(
' Extract the bounds between which the objects dimensions are needed'
)
# It is necessary for this demo that the object's length and breath is smaller than that of the chessboard
chessboard_points_cumulative_3d = np.delete(
chessboard_points_cumulative_3d, 0, 1)
self.roi_2D = get_boundary_corners_2D(chessboard_points_cumulative_3d)
print(
"Calibration completed... \nPlace the box in the field of view of the devices..."
)
def setEmitter(self):
# print('Enable the emitter of the devices')
self.device_manager.enable_emitter(True)
# print('-------Enable the emitter of the devices')
# Load the JSON settings file in order to enable High Accuracy preset for the realsense
#self.device_manager.load_settings_json("HighResHighAccuracyPreset.json")
#print(' Get the extrinsics of the device to be used later')
extrinsics_devices = self.device_manager.get_depth_to_color_extrinsics(
self.frames)
print(
' Get the calibration info as a dictionary to help with display of the measurements onto the color image instead of infra red image'
)
self.calibration_info_devices = defaultdict(list)
for calibration_info in (self.transformation_devices,
self.intrinsics_devices, extrinsics_devices):
for key, value in calibration_info.items():
self.calibration_info_devices[key].append(value)
print("CalibrasetEmittertion completed... ")
def processing(self):
frames_devices = self.device_manager.poll_frames()
print('get frames')
# Calculate the pointcloud using the depth frames from all the devices
point_cloud = calculate_cumulative_pointcloud(
frames_devices, self.calibration_info_devices, self.roi_2D)
# Get the bounding box for the pointcloud in image coordinates of the color imager
bounding_box_points_color_image, length, width, height = calculate_boundingbox_points(
point_cloud, self.calibration_info_devices)
print('길이=', length, ' 폭=', width, ' 높이', height)
# Draw the bounding box points on the color image and visualise the results
# visualise_measurements(frames_devices, bounding_box_points_color_image, length, width, height)
def closeSense(self):
self.device_manager.disable_streams()
