OpenMC Cam H7 Designed for Real-Time, Low-Power

Upgraded camera system wants to be the 'Arduino of Machine Vision'.

Kwabena Agyeman and his team from OpenMV Cam wanted a simple machine vision system that was easy to operate and powered up quickly. The design was required to be low power and run from a USB port, able to run without configuration, and a size smaller than standard single board computers. The vision is a future where cameras can be place into anything that a user desires. The group is currently running a Kickstarter campaign for the OpenMV Cam H7.

OpenMV Cam H7 is programmed in Python, and has doubled the power of its predecessor the OpenMV Cam M7. The OpenMV cams have previously been popular for color tracking, and the H7 can reach input rates of 80 frames per second. Sixteen different color thresholds are used to segment images and the H7 groups blobs of the same color in proximity, outputting box size, centroid, and orientation for each blob. Adding an adapter allows the user to track thermal images using a FLIR Lepton sensor, and other add-on modules are available for Global Shutter Sensor Support, and Convolution Neural Networks for Deep Learning.

The OpenMV Cam H7 is controlled by STM32H743VIT6 microcontroller, and the full specifications are available on the company’s webpage. The base OpenMV IDE gives users a base to start their programming right out of the box. The system is open-source and available on GitHub now, and Agyeman encourages users to create their own desktop applications. Options are available to add a clear case, the FLIR and Global Shutter modules, a prototyping shield, LCD shield, WiFi shield, shields for servos and motors, and IR / wide angle / Telephoto lenses. Some sample projects in the campaign video show a robot focusing on a ball and following it around the room, and the Donkey Self-Driving Car. This is a great example of a tool that’s already established and pushing itself to be more powerful and more useful to makers and engineers. The campaign ends on October 17, 2018 and if successful boards are currently scheduled to ship in March 2019.