Lightel DI-2000

A small Python interface for the Lightel DI-2000 USB digital inspector.

The public DI-2000 documentation describes a USB video camera but does not publish a command protocol or SDK. This package therefore uses the operating system's camera driver through OpenCV. It supports camera discovery, connection, frame acquisition, image capture, and standard UVC/OpenCV properties.

Installation

pip install .

For development:

pip install -e ".[dev]"
python -m pytest

Usage

from di2000 import Di2000, find_devices

print("Camera indexes:", find_devices())

inspector = Di2000()
inspector.connect(0)
try:
    frame = inspector.read_frame()
    saved_path = inspector.capture_image("inspection.png")
    print(saved_path)
finally:
    inspector.disconnect()

On Windows, Lightel's installed camera driver may work better through DirectShow:

import cv2
from di2000 import Di2000

with Di2000() as inspector:
    inspector.connect(0, backend=cv2.CAP_DSHOW)
    inspector.capture_image("inspection.png")

Standard camera controls can be attempted through OpenCV:

import cv2

inspector.set_autofocus(True)
inspector.set_property(cv2.CAP_PROP_BRIGHTNESS, 100)
print(inspector.get_property(cv2.CAP_PROP_BRIGHTNESS))

Vendor autofocus / capture-button protocol

Lightel has published the probe's capture button, onboard autofocus trigger, and autofocus status as a UVC Extension Unit (confirmed by USB descriptor enumeration against unit: bUnitID=3, GUID 63610682-5070-49AB-B8CC-B3855E8D221D), documented at kernel.org's uvcvideo driver doc and Microsoft's UVC 1.5 extensions doc.

inspector.trigger_autofocus()
status = inspector.get_autofocus_status()   # AutofocusStatus.RUNNING / DONE_OK / DONE_FAILED
pressed = inspector.get_capture_button_pressed()

Linux only for now, via the kernel's UVCIOC_CTRL_QUERY ioctl on /dev/videoN -- the same mechanism the kernel doc above describes, and the standard way to reach a UVC extension unit without disturbing an active V4L2 video stream. There's no macOS implementation: the built-in AVFoundation UVC driver holds the interface with no user-accessible way to detach it, so raw USB control transfers there fail at the OS level, not at the device. Windows would need a separate implementation (DirectShow/Media Foundation).

inspection.autofocus() uses trigger_autofocus()/get_autofocus_status() automatically when the camera object exposes them (see AutofocusResult.method), falling back to the generic UVC/manual-sweep heuristics otherwise.

Finding the right camera

find_devices() returns all camera indexes OpenCV can open, not only Lightel devices. Disconnect other cameras or try each returned index and inspect a frame.

Fiber-end inspection

inspection.py builds on top of a connected Di2000 to bring the fiber end into focus, classify whether the shot is front- or back-illuminated (there is no UVC control to read this from hardware, so it's detected from the image), expose for that mode, and flag dirt/debris.

from di2000 import Di2000
from inspection import inspect

with Di2000() as inspector:
    inspector.connect(0)
    result = inspect(inspector)

print(result.illumination.mode, result.defect_area_fraction, result.passed)

The individual stages (locate_fiber_end, autofocus, classify_illumination, autoexpose, detect_defects) are also usable on their own, and all operate on plain numpy frames so they can be tested without a camera.

GUI

gui.py is a small PySide6 app

pip install -e ".[gui]"
python gui.py
# or, after install: lightel-inspect-gui