Data processing and analysis

An analysis of the collected data demonstrated that the use of a system of surveillance cameras working in the Edge Computing mode is sufficient to develop complex models of time- and space-varying use of campus space by pedestrians, cyclists, cars, and delivery vehicles. Image data processed using proprietary scripts and analyzed using spatial data mining methods enable analysis of social behavior, detection of spatial dysfunctions, and indication of their location, as well as optimization of the architectural design method. Due to the speed of the change associated with the intensity of traffic, an analysis of the dynamic image may cause classification uncertainties, for example, in the recognition of objects, but this does not have a significant impact on the statistical analysis of the collected data. Similarly, due to the uneven distribution of transformation points in individual images and the associated insufficient training of neural networks in the process of conversion of image data to GIS spatial database structures, there are occasional errors in assigning object motion to the wrong basic field. However, the use of filtration of results using topographic data makes it possible to eliminate information noise and analyze the data aggregated to basic fields with a resolution appropriate for the purpose of the research. Further work on a modification of the developed technology will involve the development of proprietary software for the purpose of full automation of the processes of data processing and knowledge acquisition in the edge computing mode without the need for post-processing.