Description: The article analyzes the known methods of images thematic segmentation of onboard systems of optical-electronic observa-tion, including neural network, their main drawbacks. Thematic segmentation is understood as segmentation, feature extraction of objects of interest and semantic segmentation. The result of thematic segmentation is the separation of the image into artificial objects (objects of interest) and natural objects (background). A method is proposed for thematic segmentation of optoelectronic images based on the artificial bee colony method. It has been established that the choice of the fitness function is an important stage of the method. The requirements for the fitness function are formulated, the model images for the cases when the object and the background in the image are separated and not separated are considered. It has been established that as a fitness function it is advisable to choose a function that has the physical meaning of the sum of the variances of the segments of the segmented im-age. In the general case, an optimization problem is formulated for selecting the thematic segmentation threshold, which is to minimize the fitness function at each iteration of the iterative process of determining the threshold level.
Keywords: segmentation method, swarm method, artificial bee colony, optic-electronic image, fitness function, on-board observation system, segmentation threshold, iterative process, iterative process, object of interest, background