rahix/openflexure-microscope-server
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Move new default plugins to API submodule

Browse source
This commit is contained in:
Joel Collins 2019-12-28 18:53:10 +00:00
parent e4c6e892fa
commit 419cc411b7
4 changed files with 5 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

0
openflexure_microscope/api/default_plugins/__init__.py Normal file
View file

366
openflexure_microscope/api/default_plugins/autofocus.py Normal file
View file

@ -0,0 +1,366 @@
from openflexure_microscope.common.flask_labthings.find import find_device
from openflexure_microscope.common.flask_labthings.plugins import BasePlugin
from openflexure_microscope.common.flask_labthings.resource import Resource
from openflexure_microscope.devel import JsonResponse, request, jsonify, taskify, abort
from openflexure_microscope.utilities import set_properties
import time
import numpy as np
import threading
import logging
from scipy import ndimage
from contextlib import contextmanager
### Autofocus utilities
class JPEGSharpnessMonitor:
def __init__(self, microscope, timeout=60):
self.microscope = microscope
self.camera = microscope.camera
self.stage = microscope.stage
self.jpeg_sizes = []
self.jpeg_times = []
self.stage_positions = []
self.stage_times = []
self.stop_event = threading.Event()
self.timeout = timeout
self.keep_alive()
self.background_thread = None
def is_alive(self):
if self.background_thread is None:
return False
else:
return self.background_thread.is_alive()
def should_stop(self):
import time
return time.time() - self.kept_alive > self.timeout
def keep_alive(self):
import time
self.kept_alive = time.time()
def start(self):
"Start monitoring sharpness by looking at JPEG size"
self.background_thread = threading.Thread(target=self._measure_jpegs)
self.background_thread.start()
return self
def stop(self):
"Stop the background thread"
self.stop_event.set()
self.background_thread.join()
def _measure_jpegs(self):
"Function that runs in a background thread to record sharpness"
logging.info("Starting sharpness measurement in background thread")
self.keep_alive()
while not self.stop_event.is_set() and not self.should_stop():
self.jpeg_sizes.append(self.jpeg_size())
self.jpeg_times.append(time.time())
if self.stop_event.is_set():
logging.info("Cleanly stopped sharpness measurement in background thread")
if self.should_stop():
logging.info("Sharpness measurement timed out and has stopped")
def jpeg_size(self):
"""Return the size of a frame from the MJPEG stream"""
return len(self.camera.get_frame())
def focus_rel(self, dz, backlash=False, **kwargs):
self.keep_alive()
self.stage_times.append(time.time())
self.stage_positions.append(self.stage.position)
self.stage.move_rel([0, 0, dz], backlash=backlash, **kwargs)
self.stage_times.append(time.time())
self.stage_positions.append(self.stage.position)
i = len(self.stage_positions) - 2
return i, self.stage_positions[-1][2]
def move_data(self, istart, istop=None):
"Extract sharpness as a function of (interpolated) z"
global np, logging
if istop is None:
istop = istart + 2
jpeg_times = np.array(self.jpeg_times)
jpeg_sizes = np.array(self.jpeg_sizes)
stage_times = np.array(self.stage_times)[istart:istop]
stage_zs = np.array(self.stage_positions)[istart:istop, 2]
start = np.argmax(jpeg_times > stage_times[0])
stop = np.argmax(jpeg_times > stage_times[1])
if stop < 1:
stop = len(jpeg_times)
logging.debug("changing stop to {}".format(stop))
jpeg_times = jpeg_times[start:stop]
jpeg_zs = np.interp(jpeg_times, stage_times, stage_zs)
return jpeg_times, jpeg_zs, jpeg_sizes[start:stop]
def sharpest_z_on_move(self, index):
"""Return the z position of the sharpest image on a given move"""
jt, jz, js = self.move_data(index)
return jz[np.argmax(js)]
def data_dict(self):
"""Return the gathered data as a single convenient dictionary"""
data = {}
for k in ["jpeg_times", "jpeg_sizes", "stage_times", "stage_positions"]:
data[k] = getattr(self, k)
return data
def decimate_to(shape, image):
"""Decimate an image to reduce its size if it's too big."""
decimation = np.max(
np.ceil(np.array(image.shape, dtype=np.float)[: len(shape)] / np.array(shape))
)
return image[:: int(decimation), :: int(decimation), ...]
def sharpness_sum_lap2(rgb_image):
"""Return an image sharpness metric: sum(laplacian(image)**")"""
# image_bw=np.mean(decimate_to((1000,1000), rgb_image),2)
image_bw = np.mean(rgb_image, 2)
image_lap = ndimage.filters.laplace(image_bw)
return np.mean(image_lap.astype(np.float) ** 4)
def sharpness_edge(image):
"""Return a sharpness metric optimised for vertical lines"""
gray = np.mean(image.astype(float), 2)
n = 20
edge = np.array([[-1] * n + [1] * n])
return np.sum(
[np.sum(ndimage.filters.convolve(gray, W) ** 2) for W in [edge, edge.T]]
)
### Autofocus plugin
def measure_sharpness(microscope, metric_fn=sharpness_sum_lap2):
"""Measure the sharpness of the camera's current view."""
return metric_fn(microscope.camera.array(use_video_port=True))
def autofocus(microscope, dz, settle=0.5, metric_fn=sharpness_sum_lap2):
"""Perform a simple autofocus routine.
The stage is moved to z positions (relative to current position) in dz,
and at each position an image is captured and the sharpness function
evaulated. We then move back to the position where the sharpness was
highest. No interpolation is performed.
dz is assumed to be in ascending order (starting at -ve values)
"""
camera = microscope.camera
stage = microscope.stage
with set_properties(stage, backlash=256), stage.lock, camera.lock:
sharpnesses = []
positions = []
camera.annotate_text = ""
for _ in stage.scan_z(dz, return_to_start=False):
positions.append(stage.position[2])
time.sleep(settle)
sharpnesses.append(measure_sharpness(microscope, metric_fn))
newposition = positions[np.argmax(sharpnesses)]
stage.move_rel([0, 0, newposition - stage.position[2]])
return positions, sharpnesses
@contextmanager
def monitor_sharpness(microscope):
m = JPEGSharpnessMonitor(microscope)
m.start()
try:
yield m
finally:
m.stop()
def move_and_find_focus(microscope, dz):
"""Make a relative Z move and return the peak sharpness position"""
with monitor_sharpness(microscope) as m:
m.focus_rel(dz)
return m.sharpest_z_on_move(0)
def fast_autofocus(microscope, dz=2000, backlash=None):
"""Perform a down-up-down-up autofocus"""
with monitor_sharpness(microscope) as m:
i, z = m.focus_rel(-dz / 2)
i, z = m.focus_rel(dz)
fz = m.sharpest_z_on_move(i)
if backlash is None:
i, z = m.focus_rel(-dz) # move all the way to the start so it's consistent
else:
i, z = m.focus_rel(fz - z - backlash)
m.focus_rel(fz - z)
return m.data_dict()
def fast_up_down_up_autofocus(
microscope, dz=2000, target_z=0, initial_move_up=True, mini_backlash=150
):
"""Autofocus by measuring on the way down, and moving back up with feedback.
This autofocus method is very efficient, as it only passes the peak once.
The sequence of moves it performs is:
1. Move to the top of the range `dz/2` (can be disabled)
2. Move down by `dz` while monitoring JPEG size to find the focus.
3. Move back up to the `target_z` position, relative to the sharpest image.
4. Measure the sharpness, and compare against the curve recorded in (2) to \\
estimate how much further we need to go. Make this move, to reach our \\
target position.
Moving back to the target position in two steps allows us to correct for
backlash, by using the sharpness-vs-z curve as a rough encoder for Z.
Parameters:
dz: number of steps over which to scan (optional, default 2000)
target_z: we aim to finish at this position, relative to focus. This may
be useful if, for example, you want to acquire a stack of images in Z.
It is optional, and the default value of 0 will finish at the focus.
initial_move_up: (optional, default True) set this to `False` to move down
from the starting position. Mostly useful if you're able to combine
the initial move with something else, e.g. moving to the next scan point.
mini_backlash: (optional, default 50) is a small extra move made in step
3 to help counteract backlash. It should be small enough that you
would always expect there to be greater backlash than this. Too small
might slightly hurt accuracy, but is unlikely to be a big issue. Too big
may cause you to overshoot, which is a problem.
"""
with monitor_sharpness(microscope) as m, microscope.camera.lock:
# Ensure the MJPEG stream has started
microscope.camera.start_stream_recording()
df = dz # TODO: refactor so I actually use dz in the code below!
if initial_move_up:
m.focus_rel(df / 2)
# move down
i, z = m.focus_rel(-df)
# now inspect where the sharpest point is, and estimate the sharpness
# (JPEG size) that we should find at the start of the Z stack
jt, jz, js = m.move_data(i)
best_z = jz[np.argmax(js)]
target_s = np.interp(
[best_z + target_z], jz[::-1], js[::-1]
) # NB jz is decreasing
# now move to the start of the z stack
i, z = m.focus_rel(
best_z + target_z - z + mini_backlash
) # takes us to the start of the stack
# We've deliberately undershot - figure out how much further we should move based on the curve
current_js = m.jpeg_size()
imax = np.argmax(js) # we want to crop out just the bit below the peak
js = js[imax:] # NB z is in DECREASING order
jz = jz[imax:]
inow = np.argmax(
js < current_js
) # use the curve we recorded to estimate our position
# TODO: fancy interpolation stuff
# So, the Z position corresponding to our current sharpness value is zs[inow]
# That means we should move forwards, by best_z - zs[inow]
correction_move = best_z + target_z - jz[inow]
logging.debug(
"Fast autofocus scan: correcting backlash by moving {} steps".format(
correction_move
)
)
m.focus_rel(correction_move)
return m.data_dict()
class MeasureSharpnessAPI(Resource):
def post(self):
microscope = find_device("openflexure_microscope")
if not microscope:
abort(503, "No microscope connected. Unable to measure sharpness.")
return jsonify({"sharpness": measure_sharpness(microscope)})
class AutofocusAPI(Resource):
"""
Run a standard autofocus
"""
def post(self):
payload = JsonResponse(request)
microscope = find_device("openflexure_microscope")
if not microscope:
abort(503, "No microscope connected. Unable to autofocus.")
# Figure out the range of z values to use
dz = payload.param("dz", default=np.linspace(-300, 300, 7), convert=np.array)
if microscope.has_real_stage():
logging.info("Running autofocus...")
task = taskify(autofocus)(microscope, dz)
# return a handle on the autofocus task
return jsonify(task.state), 201
else:
abort(503, "No stage connected. Unable to autofocus.")
class FastAutofocusAPI(Resource):
"""
Run a fast autofocus
"""
def post(self):
payload = JsonResponse(request)
microscope = find_device("openflexure_microscope")
if not microscope:
abort(503, "No microscope connected. Unable to autofocus.")
# Figure out the parameters to use
dz = payload.param("dz", default=2000, convert=int)
backlash = payload.param("backlash", default=0, convert=int)
if backlash < 0:
backlash = 0
if microscope.has_real_stage():
logging.info("Running autofocus...")
task = taskify(fast_autofocus)(microscope, dz, backlash=backlash)
# return a handle on the autofocus task
return jsonify(task.state), 201
else:
abort(503, "No stage connected. Unable to autofocus.")
autofocus_plugin_v2 = BasePlugin("autofocus")
autofocus_plugin_v2.add_method(fast_autofocus, "fast_autofocus")
autofocus_plugin_v2.add_method(autofocus, "autofocus")
autofocus_plugin_v2.add_view(MeasureSharpnessAPI, "/measure_sharpness")
autofocus_plugin_v2.add_view(AutofocusAPI, "/autofocus")
autofocus_plugin_v2.register_action(AutofocusAPI)
autofocus_plugin_v2.add_view(FastAutofocusAPI, "/fast_autofocus")
autofocus_plugin_v2.register_action(FastAutofocusAPI)

404
openflexure_microscope/api/default_plugins/scan.py Normal file
View file

@ -0,0 +1,404 @@
import numpy as np
import itertools
import logging
import uuid
from typing import Tuple
from functools import reduce
from openflexure_microscope.camera.base import generate_basename
from openflexure_microscope.common.flask_labthings.find import find_device, find_plugin
from openflexure_microscope.common.flask_labthings.plugins import BasePlugin
from openflexure_microscope.devel import (
JsonResponse,
request,
jsonify,
taskify,
abort,
update_task_progress,
update_task_data,
)
from openflexure_microscope.common.flask_labthings.resource import Resource
import time
### Grid construction
def construct_grid(initial, step_sizes, n_steps, style="raster"):
"""
Given an initial position, step sizes, and number of steps,
construct a 2-dimensional list of scan x-y positions.
"""
arr = []
for i in range(n_steps[0]): # x axis
arr.append([])
for j in range(n_steps[1]): # y axis
# Create a coordinate array
coord = [initial[ax] + [i, j][ax] * step_sizes[ax] for ax in range(2)]
# Append coordinate array to position grid
arr[i].append(tuple(coord))
# Style modifiers
if style == "snake":
for i, line in enumerate(arr):
if i % 2 != 0:
line.reverse()
return arr
def flatten_grid(grid):
"""
Convert a 3D list of scan positions into a flat list
of sequential positions.
"""
grid = list(itertools.chain(*grid))
return grid
### Progress
_images_to_be_captured: int = 1
_images_captured_so_far: int = 0
def progress():
progress = (_images_captured_so_far / _images_to_be_captured) * 100
logging.info(progress)
return progress
### Capturing
def capture(
microscope,
basename,
scan_id,
temporary: bool = False,
use_video_port: bool = False,
resize: Tuple[int, int] = None,
bayer: bool = False,
metadata: dict = {},
tags: list = [],
):
# Construct a tile filename
filename = "{}_{}_{}_{}".format(basename, *microscope.stage.position)
folder = "SCAN_{}".format(basename)
# Create output object
output = microscope.camera.new_image(
temporary=temporary, filename=filename, folder=folder
)
# Capture
microscope.camera.capture(
output.file, use_video_port=use_video_port, resize=resize, bayer=bayer
)
# Affix metadata
if "scan" not in tags:
tags.append("scan")
# Inject system metadata
output.put_metadata(microscope.metadata, system=True)
# Insert custom metadata
output.put_metadata(metadata)
# Insert custom tags
output.put_tags(tags)
### Scanning
def tile(
microscope,
basename: str = None,
temporary: bool = False,
step_size: int = [2000, 1500, 100],
grid: list = [3, 3, 5],
style="raster",
autofocus_dz: int = 50,
use_video_port: bool = False,
resize: Tuple[int, int] = None,
bayer: bool = False,
fast_autofocus=False,
metadata: dict = {},
tags: list = [],
):
global _images_to_be_captured
global _images_captured_so_far
# Keep task progress
# TODO: Make this line not nasty
_images_to_be_captured = reduce((lambda x, y: x * y), grid)
_images_captured_so_far = 0
# Generate a basename if none given
if not basename:
basename = generate_basename()
# Generate a stack ID
scan_id = uuid.uuid4()
# Store initial position
initial_position = microscope.stage.position
# Add scan metadata
if "time" not in metadata:
metadata["time"] = generate_basename()
metadata.update(
{
"scan_id": scan_id,
"basename": basename,
"scan_parameters": {
"step_size": step_size,
"grid": grid,
"style": style,
"autofocus_dz": autofocus_dz,
},
}
)
# Check if autofocus is enabled
autofocus_plugin = find_plugin("autofocus")
if (
autofocus_dz
and autofocus_plugin
and microscope.has_real_stage()
and microscope.has_real_camera()
):
autofocus_enabled = True
else:
autofocus_enabled = False
z_stack_dz = (
grid[2] * step_size[2] if grid[2] > 1 else 0
) # shorthand for Z stack range
# Construct an x-y grid (worry about z later)
x_y_grid = construct_grid(initial_position, step_size[:2], grid[:2], style=style)
# Keep the initial Z position the same as our current position
next_z = initial_position[2]
if fast_autofocus: # If fast autofocus is enabled, make
next_z += autofocus_dz / 2 # sure we start from the top of the range
initial_z = next_z # Save this value for use in raster scans
# Now step through each point in the x-y coordinate array
for line in x_y_grid:
# If rastering, rather than snake (or eventually spiral)
# Return focus to initial position
if style == "raster":
next_z = initial_z
logging.debug("Returning to initial z position")
microscope.stage.move_abs(
[line[0][0], line[0][1], next_z]
) # RWB: I think this line is redundant
for x_y in line:
# Move to new grid position without changing z
logging.debug("Moving to step {}".format([x_y[0], x_y[1], next_z]))
microscope.stage.move_abs([x_y[0], x_y[1], next_z])
# Refocus
if autofocus_enabled:
if fast_autofocus:
autofocus_plugin.fast_autofocus(
dz=autofocus_dz,
target_z=-z_stack_dz / 2.0, # Finish below the focus
initial_move_up=False, # We're already at the top of the scan
)
# TODO: save the focus data for future reference? Use it for diagnostics?
else:
logging.debug("Running autofocus")
autofocus_plugin.autofocus(
range(-3 * autofocus_dz, 4 * autofocus_dz, autofocus_dz)
)
logging.debug("Finished autofocus")
time.sleep(1) # TODO: Remove
# If we're not doing a z-stack, just capture
if grid[2] <= 1:
capture(
microscope,
basename,
scan_id,
temporary=temporary,
use_video_port=use_video_port,
resize=resize,
bayer=bayer,
metadata=metadata,
tags=tags,
)
# Update task progress
_images_captured_so_far += 1
update_task_progress(progress())
else:
logging.debug("Entering z-stack")
stack(
microscope=microscope,
basename=basename,
temporary=temporary,
scan_id=scan_id,
step_size=step_size[2],
steps=grid[2],
center=not fast_autofocus, # fast_autofocus does this for us!
return_to_start=not fast_autofocus,
use_video_port=use_video_port,
resize=resize,
bayer=bayer,
metadata=metadata,
tags=tags,
)
# Make sure we use our current best estimate of focus (i.e. the current position) next point
next_z = microscope.stage.position[2]
if fast_autofocus:
next_z += (
autofocus_dz / 2
) # Fast autofocus requires us to start at the top of the range
if grid[2] > 1:
next_z -= int(
grid[2] / 2.0 * step_size[2]
) # Z stacking means we're higher up to start with
logging.debug("Returning to {}".format(initial_position))
microscope.stage.move_abs(initial_position)
def stack(
microscope,
basename: str = None,
temporary: bool = False,
scan_id: str = None,
step_size: int = 100,
steps: int = 5,
center: bool = True,
return_to_start: bool = True,
use_video_port: bool = False,
resize: Tuple[int, int] = None,
bayer: bool = False,
metadata: dict = {},
tags: list = [],
):
global _images_captured_so_far
# Generate a basename if none given
if not basename:
basename = generate_basename()
# Generate a stack ID
if not scan_id:
scan_id = uuid.uuid4()
# Add scan metadata
if not "time" in metadata:
metadata["time"] = generate_basename()
# Store initial position
initial_position = microscope.stage.position
with microscope.lock:
# Move to center scan
if center:
logging.debug("Moving to starting position")
microscope.stage.move_rel([0, 0, int((-step_size * steps) / 2)])
for i in range(steps):
time.sleep(0.1)
logging.debug("Capturing...")
capture(
microscope,
basename,
scan_id,
temporary=temporary,
use_video_port=use_video_port,
resize=resize,
bayer=bayer,
metadata=metadata,
tags=tags,
)
# Update task progress
_images_captured_so_far += 1
update_task_progress(progress())
if i != steps - 1:
logging.debug("Moving z by {}".format(step_size))
microscope.stage.move_rel([0, 0, step_size])
if return_to_start:
logging.debug("Returning to {}".format(initial_position))
microscope.stage.move_abs(initial_position)
### Web views
class TileScanAPI(Resource):
def post(self):
payload = JsonResponse(request)
microscope = find_device("openflexure_microscope")
if not microscope:
abort(503, "No microscope connected. Unable to autofocus.")
# Get params
filename = payload.param("filename")
temporary = payload.param("temporary", default=False, convert=bool)
step_size = payload.param("step_size", default=[2000, 1500, 100], convert=list)
step_size = [int(i) for i in step_size]
grid = payload.param("grid", default=[3, 3, 5], convert=list)
grid = [int(i) for i in grid]
style = payload.param("style", default="raster", convert=str)
autofocus_dz = payload.param("autofocus_dz", default=50, convert=int)
fast_autofocus = payload.param("fast_autofocus", default=False, convert=bool)
use_video_port = payload.param("use_video_port", default=True, convert=bool)
resize = payload.param("size", default=None)
if resize:
if ("width" in resize) and ("height" in resize):
resize = (
int(resize["width"]),
int(resize["height"]),
) # Convert dict to tuple
else:
abort(404)
bayer = payload.param("bayer", default=False, convert=bool)
metadata = payload.param("metadata", default={}, convert=dict)
tags = payload.param("tags", default=[], convert=list)
logging.info("Running tile scan...")
task = taskify(tile)(
microscope,
basename=filename,
temporary=temporary,
step_size=step_size,
grid=grid,
style=style,
autofocus_dz=autofocus_dz,
use_video_port=use_video_port,
resize=resize,
bayer=bayer,
fast_autofocus=fast_autofocus,
metadata=metadata,
tags=tags,
)
# return a handle on the scan task
return jsonify(task.state), 201
scan_plugin_v2 = BasePlugin("scan")
scan_plugin_v2.add_view(TileScanAPI, "/tile")
scan_plugin_v2.register_action(TileScanAPI)