Digit OCR on stm or esp from onboard camera

Embedded Systems & Microcontrollers — incorrectly specified categories?

Scansione di un'immagine, riconoscimento delle posizioni delle cifre, taglio dell'immagine a una cifra, ridimensionamento dell'immagine a 2 colori, OCR digitale, trasmissione dei risultati al PC via e-mail o RF diretta.

Update #1 from 29 February 2020

In this project you must do this tasks:

- scan an image (I will give you some images) in these images there are numbers

- recognize number positions

- cut the image to leave only a digit

- resampling image to 2 color depth

- OCR the image to recover the real number in it in BCD

- send the BCD and the cutted image files to a remote pc.

You can use any low power, low cost device for this purpose.

You can't use pc as recognition server. OCR must be done on microprocessor embedded device.

There are more tasks to do that I don't explain now.

Update #2 from 1 March 2020

Due to ultra low power performances I preffer to use STM32 instead of ESP32.

My informations are that STM32 (STM32 L4+) absorbs 1uA while ESP32 absorbs 170uA in ultra low power mode.

I need to use it in battery operated long time installation and 170uA is too much to ensure 10 yrs life.

Proposals 5

Aleksandr L.

9 0

Projects 9
Rating -
Rating 644

Budget: 65000 UAH Deadline: 21 days

Hello
1. STM32L476 (420 nA Standby mode with RTC to wake up regulary)
2. OmniVision camera (2640/7725 etc.. with wide angles lenses) ~ 25mA
3. OCR digits recognition algorithm -> writing result to flash mem
4. mosfet power control of Camera for standby mode
5. usb/serial or other interface to transmit results
6. 1..5 buttons/ jumpers/leds

Approximate work and budget:

Milestone A) Virtual simulation of the electrical circuit + complete code writing + modular prototyping ~2500 $

Milestone B) PCB ~ 1000$

the home laboratory has all the necessary equipment (including various versions of MCUs and discovery boards), except cameras with the required specific optics of short focus, so additional budget for fast delivery of camera modules is required.

Artur Shidlovsky

0 0

Projects -
Rating -
Rating 180

Budget: 200 UAH Deadline: 1 day

We can complete this project using a neural network to recognize numbers on board STM MCU. e.g. Tensor flow. We will determine the cost and term after receiving the technical specifications.

Sergey Ivanov

36 0

Projects 37
Rating -
Rating 2 751

Budget: 75000 UAH Deadline: 25 days

I became interested in your project. Ready for cooperation. I see the implementation of the project on low-power STM processors using opencv libraries