Application of image recognition technology based on the theory of machine learning – traditional water meter with remote meter reading function
Market Evolution
Due to the rapid development in the artificial intelligence field in recent years, various industries have started introducing machine learning technology for application development. The machine learning interpretation field is mainly divided into four types of data input, including video, image, voice, and vibration, in which image recognition technology is often used in fields such as license plate recognition and water meter fee identification, etc.
Traditional water consumes manpower for reading the meters
The calculation of the fees with traditional water meters requires the use of large numbers of manpower to read the meters for water usage, especially in remote areas where maintenance is even more difficult, and water companies need to invest significant amounts of personnel costs for this. With the emerging of IoT and machine learning technologies, the fee information on the water meter is first identified and then transmitted backstage through wireless networks to reduce the consumption of personnel costs. This has become an emerging practice. However, replacing water meters comprehensively takes a very long time, not to mention the required costs; therefore, alternative solutions were developed.
Numeric character recognition technology using wireless camera modules and machine learning
In order to achieve the goal described above, customers are provided with an option to implement machine learning recognition technology quickly and easily. Nuvoton released smart numeric character recognition solutions based on the NuMicro M48xxGCAE series. Install CMOS camera modules on existing traditional water meters to regularly take photos of the water meter fees, use image recognition technology to identify the numbers, and transmit the water fee information using wireless modules.
NuMicro M48xxGCAE series provides rich peripheral interfaces, including 1 CMOS camera interface, 1 full-speed USB, 3 sets of CAN buses, 8 UARTs, 1 ISO7816/UART, 3 I2C, 3 SPI/I2S, 2QSPI, 1 I2S, 1 EBI, 24-channel PWM, 2 12-bit 5MSPS ADC with up to 16 channels, 1 12-bit 1MSPS DAC and 2 analog comparators.
The following figure is the architectural diagram for a smart numeric character recognition water meter; Nuvoton not only provides the M48xxGCAE series development platform, but it also provides sample codes for numeric character recognition technology based on machine learning theory, able to reduce the difficulty for development and significantly save customers large amounts of development time.
If you want to learn more about the Nuvoton M480 series, please refer to the following link to visit the official website of Nuvoton
https://www.nuvoton.com/products/microcontrollers/arm-cortex-m4-mcus/index.html
If you want to evaluate the NuMaker-M483KG development platform for projects, it can be purchased from Nuvoton’s online store or Nuvoton’s flagship store on Tmall.