Today according to UVS International (the leading international association of pilotless systems) unmanned aerial vehicles (UAVs) make in 52 countries of the world. Tens big enterprises and small firms compete in this market.
One of the major elements for the UAV is the system of navigation. Now there is a set of the navigation systems working independently from each other, including, satellite navigation, inertial navigation systems, navigation with use of data of a video series. The systems transforming data of a video series to so-called contour maps on which the UAV has opportunity to be guided are developed. But, as often we have no physical capacity of storage onboard the UAV of large volume of data, we decided to go further away and instead of creation of the card to create a database of specific objects.
Development of algorithm of navigation according to a video series for return of the UAV to the place of basing in case of the emergency which caused malfunctions in systems of navigation of the aircraft or a blackout of its systems of navigation was the purpose of this work.
At creation of a database of specific objects we face a problem of preliminary processing of the image. It consists in allocation of objects necessary for us in a shot or, in other words, underlinings of contours of objects. There is a set of versions of solutions of this problem. On the solution chosen by us the first step is segmentation of the image. [1]
In our task we will use splitting the image into uniform areas. Namely, we will divide the image into 25 fixed parts equal by the size each other. It allows to solve the problem connected with various levels of brightness in various areas of the image. Besides, such segmentation allows to use further information on location of a segment with object in a shot for navigation of the UAV.
The following step when processing the image is the binarization of its segments with allocation of contours. There is a set of methods of a binarization, but not all conform to our requirements, namely, to the speed and quality of underlining of contours. After the analysis of the existing methods of binarization the decision was made that the most optimum for our task is Otsu's method. [2] Its main advantages are simplicity of realization, high level of adaptation to different images by a choice of the most optimum threshold, fast time of performance. Shortcomings of a method are blocked by preliminary segmentation of the image.
After we allocated a contour of objects in a shot, we need to define what of contours as much as possible conforms to our requirements. We accept as the best the contour with the longest border which is entirely within a segment.
Having allocated the best contour, so, and object in a segment, we are obliged to write down parameters of this contour in a database. Resistance to scaling has to be one of criteria of a choice of parameters of objects. Besides, it is obvious that at shift of object in a shot its parameters shouldn't change. So, there has to be no binding to coordinates in the image. In other words, parameters of object have to be steady against linear movements of object. Object parameters also have to be steady against different turns of system of coordinates. We found four parameters conforming to these requirements.
Except the parameters necessary for recognition of object, we need the parameters necessary for making decision on location of the UAV. A segment arrangement in a shot, the direction of the movement UAV concerning a shot, indications of inertial (in particular, the magnetometer) and navigation satellite systems, attached to shots will be such parameters.
Thus, we have sets from eight parameters for each object that we can write down in a database. Actually, the dynamic array by the 8xn size, where n – amount of the written-down objects is stored in cells of memory. Recognition is conducted only in the first four parameters from eight. Other parameters are used for making decision on the current location of the UAV in case of emergencies.
Record of objects in a database has to be kept constantly when flying the UAV in the set point. In case of break of communication, existence of hindrances, passes losses of one of the main UAVs navigation systems into emergency operation. The system of navigation on a collected database joins. The UAV is unwrapped on 180 degrees and makes preliminary processing of the image.
Further search of objects in algorithm of making decisions on existence of object in a database is carried out. In case of detection of the object existing in base the algorithm will use 4 additional parameters of object which are stored in a database for determination of the current location.
The developed navigation system is a good additional method for determination of location of the UAV. Using this system as the additional level of protection against emergency situations, we have an opportunity considerably to increase probability of not crash landing of the UAV and consequently, and to keep the expensive equipment.