Artificial Retina Skunkworks
An "Artificial Retina Skunkworks" project started on April 2nd, 1999. The reason for the project was it would increase the sales of the AR chip (and other Mitsubishi Electric chips such as the M32 and M16 series) by developing prototype systems utilizing these chips. Prototypes developed: 1. Elevator People Counter - counting people at elevator entrance. 2. Proximity Detector - measures distance to the moving object based on how far the object is out of focus of the camera. 3. Camera Motion Gauge - translation and angular camera motion measurement using built-in features of AR chip to speed up the calculations. 4. LegCam - measuring the distance between vending machine and people based on monitoring people's legs activity. 5. Image Acquisition Toolkits: Matrox, WebCamII, Artificial Retina and Pamette Stereo Board. 6. Java-based graphical user interface for stereo video image acquisition and analysis for automatic detection of blind spot of a car.
Background & Objective: The goal is to locate specific opportunities in the North American market that can be satisfied by vision-enabled ubiquitous computing devices; and creating prototype systems.
Technical Discussion: The developed prototypes utilize image-processing-on-chip feature of AR sensor. They perform the very simple task: evaluate simple human activity - walking, wheeling, staying, gathering . People Counter prototype could be used for counting people in lines, LegCam - to count customers who did not buy the product but was close enough to be considered as "looking to buy". Camera Motion Gauge could be used for such application like optical mouse. Java-based GUI for the stereo imaging algorithm development for the car side blind spot detection system allowed viewing and manipulation of up to 12 images(multiple stereo), flexible positioning of images inside the GUI, ability to work with images sequences as well as single images (acquisition, viewing, analysis, storage), ability to draw geometry (points, lines and polygons) over still and video images in real-time, real-time report about pixel properties under the cursor.
Technology Area: Computer Vision
Modification Date: October 3, 2007