[ home / rules / faq / search ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM / ufo / 420 ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]

/tech/ - Technology

"Technology reveals the active relation of man to nature" - Karl Marx
Name
Options
Subject
Comment
Flag
File
Embed
Password(For file deletion.)

Check out our new store at shop.leftypol.org!


File: 1787846976326-7.png (1.73 MB, 1741x2048, 1787283890236181.png)

 

you guys are not my political equals. i am vastly different in my disposition. that isn't what I'm here to discuss. I'm here for results against a surveillance state and to build a better future.
i am looking to build a low-cost sacrificial CMOS sensor test rig. interested in characterizing how permanent sensor degradation propagates through an image pipeline and eventually causes OCR failure. not primarily interested in temporary saturation/blinding. the question is where the functional threshold actually occurs after persistent sensor damage accumulates.
>Hardware
Raspberry Pi 4 as controller and replaceable CSI camera modules. IMX477 HQ Camera as primary test sensor IMX296 global-shutter comparison. IMX219 as cheap disposable baseline. interchangeable optics, fixed hardware and software baseline.
>Optical source
controlled, wavelength-defined laser source. output will be measured and recorded to construct a reproducible guide. fixed image sensor mounting and consistent exposure intensity and angling analagous to real world conditions. treat each camera module as a sacrificial test article. the goal is the destruction of optical recognition capabilities in a reproducible and on-the-go formula.
>Software
Raspberry Pi OS/libcamera OpenCV + NumPy Tesseract + PaddleOCR RAW + processed-frame capture automated OCR against a standardized text target
>Goal
log dead/stuck pixel count, persistent defect distribution saturation/clipping noise/SNR contrast and edge degradation OCR confidence character error rate / word error rate
>tl;dr
baseline sensor, controlled exposure, capture RAW + processed frame, run OCR, map persistent defects, repeat with fresh sensor
The interesting result is the curve:
sensor degradation → image-quality degradation → OCR error rate
rather than simply declaring the camera “dead” when OCR first fails.
Looking for recommendations on Pi-compatible sensors that would make useful analogues for higher-end commercial OCR/machine-vision cameras, particularly sensors with RAW access and well-documented characteristics.

not my problem

>>34213
it is our collective problem. do you enjoy my language, comrade?


Unique IPs: 2

[Return][Go to top] [Catalog] | [Home][Post a Reply]
Delete Post [ ]
[ home / rules / faq / search ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / lgbt / latam / hobby / tech / games / anime / music / draw / AKM / ufo / 420 ] [ meta ] [ wiki / shop / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]