This file type includes high resolution graphics and schematics when applicable.
Many machine sensors like accelerometers and gyroscopes have fallen in price because of the popularity of smart phones. Low-cost cameras and optical sensors have also benefited from smart phones with camera sensors for photography or video conferencing.
Related Articles
- Will Embedded Robots Become The "Erector Set" Of The 2000s?
- Highlights from the Embedded Vision Summit
Almost a decade ago my daughter built a number of robotic science fair projects with a CMUcam that was developed at Carnegie Mellon University. It used an RGB camera. She turned that information into HSI (hue, saturation, intensity) to more readily track changes in the image. Back then the technology allowed results measured in seconds per frame.
I helped turn her Java code into assembler and put it on the CMUcam, replacing CMU’s code. The resolution was still low but results were available in tens of frames per second.
Fast forward to the Pixy CMUcam 5, which also was developed at CMU and Charmed Labs (Fig. 1). The Omnivision OV9715 RGB camera is a supported automotive camera, so it will be available for a while. It can capture a 1280- by 800-pixel image, but the system uses a lower resolution so it can handle a higher frame rate.
The board has a 204-MHz NXP LPC4330 microcontroller (see “New Platform Approaches Deliver Top Digital Designs In 2010”). The micro has a Cortex-M0 and a Cortex-M4 core. In an interesting twist, the Cortex-M0 handles the camera reading and processing a frame at a time. It converts the RGB camera information so the Cortex-M4 can perform the frame analysis. The micro has 1 Mbyte of flash and 256 kbytes of RAM. It also has a high-speed quad-SPI (QSPI) that can handle high-speed serial flash memory.
The module exposes a USB, serial port, SPI, and I2C interface. A 10-pin header is designed to link the system to an Arduino-compatible host. The USB interface is handy for PCs. The USB port can supply power. Two ports enable the LPC4330 to control positioning servos.
Kickstarter was used to start off the project. The module alone is $59. Different versions are available including one that comes with servos. The open-source software is designed to be augmented. Eventually the system might include a Python interpreter, allowing even more image processing chores to be offloaded.
Check out the Pixy Kickstarter page if you want to get one soon. They already have enough support to start building. I've already put my bid in for one. I would love to see an FPGA instead of a micro but it would easily double the price, make it a lot harder to use but it would be fast.
The software can capture the color of objects. It can also recognize color codes (CC). A CC is two or more adjacent blobs of color. CCs can be tracked automatically. They also could be used to identify a particular type of object like a charging station or a goal.
Illuminating Fingers
Leap Motion’s Controller also handles object recognition but is designed specifically to recognize hand and finger gestures (Fig. 2). It usually sits in front of a keyboard or display. The technology could be incorporated directly into something like a Mac or Windows laptop or a desktop display.
The Controller uses infrared emitters and a pair of cameras to recognize a user’s fingers and hands. It handles gesture recognition and delivers the information via the USB connection. It can be tied into applications to handle 3D positioning and gestures like pinch/expand zooming. Scrolling now is possible with a swipe.
The Controller is designed to be used with PC applications. The Pixy is designed as a low-cost sensor for robotics. Both should lead to some interesting applications.