Abstract: Modern systems of computer vision use intelligent algorithms that solve a wide class of problems from simple text recognition to complex systems of spatial orientation. One of the main problems for developers of such systems is in selection of unique attributes which remain invariant to various kinds of transformations. The article presents a comparative analysis of methods of detection and spatial localization of groups of point objects. The reviewed methods are compared by the performance and efficiency at specified dimensions. As of today there are no universal approaches to determine of such attributes, and its’ selection depends on the context of the problem being solved and on the registered conditions of observation. Various kinds of descriptors such as points, lines, angles and geometric primitives can be selected as dominating attributes. The authors study  algorithms for detection of groups of point objects based on the minimum spanning tree (MST) and using a model of associated continuous image (ACI).