Raspberry Pi Python OCR Optimization

Hello,

I am working on a lightweight computer vision prototype using a Raspberry Pi Zero 2 W.

The system captures images via a Raspberry Pi camera module and processes them locally (offline) using Python and Tesseract OCR.

The current setup is functional, but recognition accuracy is inconsistent under real-world conditions.
_____
Current setup:

Raspberry Pi Zero 2 W
Raspberry Pi Camera Module (IMX708)
Python (3.x)
Tesseract OCR (v5.x)
ImageMagick / basic preprocessing
__________
The OCR output varies depending on:

lighting conditions
background contrast
font variations
distance / scaling

The system needs to reliably extract numeric values from images in a structured format.
like 1,99€ / 15,56€ Like this. NOT BARCODES! ONLY THE PRICES !
__________
Task:

I am looking for a developer with experience in OCR and image processing to:

Improve preprocessing pipeline
(grayscale, thresholding, resizing, sharpening, ROI selection)
Optimize OCR configuration
(character whitelist, segmentation mode, tuning for numeric input)
Increase detection reliability across varying conditions
Propose a clean, efficient pipeline suitable for low-power hardware
_____________
Technical requirements:

Focus strictly on numeric recognition (0–9, comma, dot)
Handle different visual styles and contrasts
Must run efficiently on Raspberry Pi (low CPU usage)
No cloud or API solutions – fully offline
________________
Nice to have:

Experience with OpenCV
Experience optimizing OCR on embedded systems
Suggestions for performance improvements
____________________
What I provide:

Sample images
Existing Python code
Current preprocessing workflow
________________
Expected result:

A significantly improved and stable OCR pipeline for numeric detection under varying real-world conditions.
__________________
Budget:
Fixed price – open to reasonable offers depending on experience
__________________

The budget for this task is set between 80–150 EUR, depending on experience and approach.

High-quality work is highly valued. If the results exceed expectations, I am open to providing a performance-based bonus and the possibility of ongoing collaboration with higher compensation.