Image Recognition Technology plus Wireless Camera Module
This solution can identify 10 digits images (0~9); equipped with Nuvoton NuMicro® M480 series microcontroller based on the arm Cortex®-M4F core that has a basic frequency of 192 MHz, it can quickly run user algorithms and it is also equipped with floating-point unit (FPU) and digital signal processing instructions (DSP), effectively improving the execution performance of the machine learning neural network algorithm significantly. Due to the high-performance core execution speed and a maximum flash memory of 512KB, as well as the maximum built-in SRAM of 160 KB large capacity memory, software can be used to drive CMOS sensors directly; its image frame rate can reach a maximum of 30 FPS and can be applied for capturing numbers on water, electricity and gas meters and then perform recognition. The identified data can be sent to the cloud through Wi-Fi or NB-IoT, achieving the function of remote meter reading. The built-in EBI interface x80 or SPI interface can be used to connect external LCD screens to display images in real-time. The SDHC interface can be used to install an SD card to store the photos that were taken and related information. In addition, the OS has passed AliOS, Amazon FreeRTOS, and Mbed OS certifications, and it has multiple UART interfaces able to connect different wireless transmission modules simultaneously to achieve remote transmission functions. What does Nuvoton provide?
Nuvoton provides a complete set of image recognition development package, download URL of related sources is as follows:
https://www.nuvoton.com/hq/applications ... _locale=en
Contents include:
- How to connect CMOS sensor expansion module to capture image data
- How to use python to write neural network training codes on PC
- How to quantify the trained models (training parameters) and extract them
- How to use CMSIS-NN library to write the neural network architecture on the microcontroller to read quantified parameters to perform image recognition