nm-lcd-check

Overview

nm-lcd-check verifies the status of segments on an LCD screen. Given photos of a device, it locates the LCD screen in each image and checks whether each segment is in the expected state (on or off).

Installation

git clone [email protected]:NextCenturyMeters/nm-lcd-check.git
cd nm-lcd-check
npm install

How it works

Verification requires three types of images for each device model:

• Template image — a tight crop of just the LCD screen with all segments off. Used to locate the screen's position in every subsequent photo.
• All-off image — a full, uncropped photo of the device with all segments off. The template is matched against this image to find the screen's coordinates, which are then applied to test images.
• Test images — full, uncropped photos of the device in various states to verify against expected segment states.

Workflow

Step 1 — Take your photos

Take a full photo of the device with all segments off (the all-off image), crop a tight version of just the LCD area from it (the template image), and take full photos of the device in the states you want to verify (test images).

Step 2 — Label segments with the pixel picker

The pixel picker is a Python tool that lets you draw rectangles over segments on the LCD and label them. It outputs a JSON file in the correct format to pass directly as pixelData to the verifier.

Requirements: Python 3.9+, with matplotlib, numpy, and opencv-python installed.

cd pixelPicker
python -m venv venv
source venv/bin/activate      # Windows: venv\Scripts\activate
pip install matplotlib numpy opencv-python

Open picker.py and set imagePath to your template image path, then run:

python picker.py

Draw a rectangle over each segment, type a label, and press Enter to save it. Green rectangles are segments expected to be on; regions with little to no pixel activity are automatically classified as expected-blank. Click Complete when done — this writes picked.json.

To visually verify your selections, run placer.py (set imagePath to a debug/cropped image and inputFile to picked.json).

Step 3 — Run verification

const { LCDVerifier } = require('@nextmeters/lcd-check');

const verifier = new LCDVerifier();
const results = await verifier.verify({
  templateImagePath: 'path/to/template.png',
  allOffImagePath:   'path/to/allOff.png',
  tests: [
    {
      imagePath: 'path/to/testImage.png',
      pixelData: require('./picked.json'),
    },
  ],
  debugPath: 'path/to/debugOutput', // optional
});

// results is an array — one entry per test image
results.forEach(result => {
  console.log(result.imagePath, result.passed);
  if (!result.passed) {
    console.log(result.problemRegions); // array of segments that failed
  }
});

Each entry in results has:

• imagePath — the path of the image that was tested
• passed — true if all segments matched expectations
• problemRegions — array of segments that did not match, each with a problemMessage and the region's coordinates

Debugging

Passing debugPath to verify() writes intermediate images to that directory:

• allOff.png — the cropped baseline (all segments off)
• cropped.png — the cropped test image
• diff.png — pixel diff between baseline and test
• diff-gray.png — grayscale diff used for segment analysis