Machine Vision System
On the global market of ADAS systems and autopilot vehicles there is a problem of obtaining correctly read information from video devices during bad weather conditions (rain, snow).
It reduces effectiveness of the system when it is especially needed. Compensation is achieved by a combination of expensive radars and increase in computer power which takes valuable transport space or space of a standalone device by reducing the dynamic and agile performance.
A set of algorithms for noise removal from videos in real time (40fps) mode was created. It works on cheap (40 USD) computing devices that can significantly reduce the cost in comparison with analogues which work on graphics cards priced from $ 100. Removing droplets in space and replacing them with a uniform background can improve the accuracy of detection that provides increased safety on the road.
Solution and Statement details
The software reduces by half the amount of noise in the image. The software has reduced the time for detection of the desired object using the same computing power. The main achievement was the speed of the system, which is a hundred times faster than any analog on the market.