OpenMV mit i.MX RT

Image processing applications with machine learning and neural networks are very computationally intensive and consume a lot of energy. New ARM Cortex M7 microcontrollers are capable of running these demanding applications. The new microcontroller family i.MX RT by NXP with ARM Cortex M7 Core allows clock frequencies up to 1GHz. To evaluate the i.MX RT controllers, the open source image processing framework OpenMV was ported. Performance and power measurements were performed, the complex memory architecture was analyzed and comparisons were made to OpenMV on the STM32H7. Since OpenMV is based on MicroPython, it was ported to the i.MX RT family using the NXP SDK. For the development platform (Seeed-Arch-Mix Board with i.MX RT1052) a PCB was designed to connect various hardware extensions. A complete OpenMV port for the i.MX RT1052 was created. Different cameras are supported, even thermal images can be evaluated. OpenMV on the i.MX RT1052 consumes 0.5W and has a performance comparable to the STM32H7. The developed hardware and software enable a comprehensive OpenMV platform based on the i.MX RT1052.

Studienbetreuer: Erich Styger