The paper proposes a method of inter-frame data alignment, built on the operations of pre-processing and selection of object contours and based on the following basic operations: reading data from cameras, determining parameters and dimensions; solving the problem of contouring objects and simplifying them to form a mask for subsequent processing; filtering, using a multi-criteria method based on a combination of criteria, the standard deviation of the final differences of the generated estimates and the first difference of the deviation from the input implementation; the use of NS to search for animals or/and humans; parallel detection of objects located in the range of 34°C-42°C degrees Celsius (95°F-107.6°F); the use of NN to determine the type of object by analyzing the thermal portrait; use of a mask of live objects (visual data) and thermal imaging parameters as alignment centers; determining a thermal image transformation matrix; stitching a pair of images. On a set of test sequences, studies have been carried out to increase the degree of detection of living objects when a thermal imaging camera is included in the data detector. The studies were carried out on a pair of cameras of the visible spectrum (1920x1080, RGB, 8 bit) and thermal (240x180, RGB, 8 bit).
|