eng

Artificial Intelligence

31 october 2019

Much has been said about Artificial Intelligence (AI), how cool it is, what tasks it can "independently" solve. And what is “under the hood” of this trend of our time.

Simply put, the essence of artificial intelligence comes down to decision-making by the algorithm, depending on many environmental factors. At the same time, AI will not execute a program with a previously known result, but will survive and adapt to the environment, interact with the world. It is rare when it can be predicted the actual outcome of events in the life cycle of such an algorithm. It’s not easy to get practical benefits for a person from just such an AI. It may get out of control or you will spend more time understanding the logic of his decisions than solving the problems themselves.

Another thing is AI elements that can be useful for quick decision making when working with large amounts of information. Such software modules, for the most part, are based on trained “neural networks”. The essence of such training is reduced to thousands of runs through the algorithm of a digital image of the detected object. It can be a photograph, a frame of a video series, an audio fragment, the harmonic of a radio frequency, gas analyzer data ... anything you can digitize. After painstaking training, the neural network produces certain signs that are repeated for objects of the same type. The program with a high degree of probability detects an object by finding these signs in the received amount of information. The more objects we train the system, the more qualitatively the surrounding world will be recognized.

The main purpose of Spaitech company is to ensure global security by creating unique unmanned solutions. Following its goal, the company develops and applies technologies in its products that keep pace with the times or sometimes half a step ahead of it. An intelligent sensor for the "Digital Eye" UAV with an integrated neural network in full functionality will be presented shortly. At the moment, the universal flying platform Windhower is the carrier of this payload. The idea of developers in creating a tool to help drone operators and the maximum automation of information received.

It will look as follows. Before the flight, the operator indicates what kind of work he needs to carry out - patrolling, reconnaissance, search, target adjustment, cartography. If this is the last option, then there’s no point in helping him, this flight will take place in auto mode. And in other modes, help will not hurt the operator. During the flight, "Digital Eye" will detect people and equipment using the received video from the day camera and the thermal imager, classify them and highlight them for the operator on the same video and map with the coordinates of the detected object. In addition, an analytical collection of information will be conducted on how many people and equipment were discovered during the mission. The more such devices will be in operation, the more information will be available for subsequent analytics and until the neural network is trained in new objects of observation.

Artificial Intelligence
Have questions?