The work of the Visensi's team has begun as participation in security/defense projects of determining intruder presence in an restricted area. We used there acoustic, seismic, IR, microwave, X-wave signature of objects and smart algorithms for recognizing and characterization of the objects in such types of data.
We continued with image recognition through mesh of networked intelligent cameras, were each recognizes truck from tank, or bike from car, or human from animal. The problems of background subtraction where solved continuously and when moving background (trees and grass) were successfully ignored, moving camera applications where also explored.
Visensi Ltd. is a start-up, founded by scientists from BAS.bg and focused on sensors and their applications in practice. We concentrate into integrating sensors in different microprocessor systems, taking custom demands and also we sell solutions to developers and general public on the open market. As far as the sensors have their dark side (ask Edward Snowden) our moto is "sensors with privacy" - we try to integrate sensors with little room to have somebody using them for unappropriate access.
Apart from customer development orders, the first open market product, now we launch, is the Vifff-024 Beagleboard compatible camera board, which uses the very sensitive industrial grade CMOS image sensor of Aptina - MT9V024.
ViFFF-024 camera board kit. The aim of this product is to give you the chance to use the excellent MT9V024 Aptina sensor on OMAP platforms, like Beagleboard XM. The advantages of this highly configurable sensor are its sensitivity which allows for video object recognition down to quarter moon light, a very comprehensive set of control options, a high dynamic range (up to 100dB), a global shutter which avoids some video artifacts, a 10 bit pixel interface with 12 bit prior ADC discretization. The pixel size is very large (6 x 6 microns) which is the source of its highly sensitive capabilities, the very low dark current and the need for deep, 4096 (12-bit) discretization value of the pixel. Therefore this 752x480 pixel sensor is oriented toward industrial, CCTV and other very professional applications. Our CS mount holder allows to you to use different types of lenses according your possible applications.
Our board gives you the possibility to begin with the development of products using this sensor, by having an ease of use and quick take off by joining our image sensor board to your Beagle board. We give you an OpenCV primer code and also a short presentation and demonstration of the sensor sensitivity and configurability potential. You can sell it also as a part of a final product (in this case we include a cable to link it and we made price discounts proportional to the quantities) or ask us to produce a version of it, which fulfills your specific requirements.
An example that demonstrates linear Hough transform used as feature extraction technique in order to determine the roadway boundaries.
This example demonstrates template matching used for vehicle detection.