Description: In the article work to improve the quality of decoding the image obtained from the on-board system of optic-electronic observation proposed to conduct a thematic segmentation of a large-scale sequence of the same image. Well-known approaches to the analysis of large-scale information have been analyzed and it has been established that in the case where small objects with sharp boundaries are present in the image, the exact definition of the boundaries of these objects is complicated. As a method of thematic segmentation of the large-scale sequence of images, it is proposed to select a swarm method intelligence (artificial bee colony). The mathematical formulation of the segmentation of a large-scale sequence of images is presented. It has been established that segmentation threshold is chosen as the optimized parameter, and optimization is to minimize or maximize the target function. As a fitness function, the amount of dispersion is selected within each image segment, while optimizing the segmentation process and defining the threshold is to minimize the fitness function. It has been established that the segmentation process is an iterative process and the results of the iterative process of determining the threshold at some stages of the iteration are given. At each iteration stage, for each agent (bee) the value of the target function is calculated, a comparison of these values is made, a minimum value choice that corresponds to the optimal threshold according to which the sequence of images is segmented. It has been established that the improved swarm method (artificial bee colony) for the thematic segmentation of the large-scale sequence of images obtained from the on-board system of optic-electronic observation is as follows: carrying out each of the images of a large-scale sequence using the artificial bee colony, which consists in determining the optimal threshold, the value of which corresponds to the minimum of the target function; re-scaling of each segmented image to the original scale; calculation of image-filter, calculated as a result of averaging of scalable images for each pixel; finding the pixel product of the image of the filter and the original image with a scale factor (t = 1) and making such a pixel product as a segmented image.
Keywords: segmentation method, swarm method, artificial bee colony, optic-electronic image, multi-scale sequence, fitness function, on-board surveillance system, segmentation threshold, iterative process