REVIEW OF TECHNOLOGIES FOR STORING AND ANALYZING DATA FROM SATELLITE IMAGES AND AERIAL PHOTOGRAPHY

Abstract

An overview of current technologies for storing and analyzing data from satellite imagery and aerial photography is provided. Currently, the processing and aggregation industry is actively developing: new analytical methods are being added, and fundamentally new approaches are being created. To evaluate existing technologies, the article briefly describes the main products on the market, such as EOSDA, Maxar, Google Earth Engine, ArcGIS, and others. To evaluate the strengths and weaknesses, a comparative table was compiled taking into account the factors of autonomy and openness to modifications. The autonomy of geospatial data aggregation and processing is important not only in terms of dependence on third-party computing resources, but is also a key factor for implementation in critical industries. In addition, independence and the ability to be deployed on client hardware require analytical algorithms to use computing resources more economically, which in turn correlates with the quality of data aggregation approaches. Therefore, a key direction on the way to system autonomy is the ability to get rid of unnecessary information before it is processed by more demanding algorithms for searching and identifying objects or spatial patterns. That is why the authors proposed an approach to optimize processing and storage, which will simplify access to analytics results and overall performance. This approach can significantly reduce data storage and access time, while ensuring high accuracy and relevance of information. A method for aggregating aerial photography data into semi-automated panoramic images that remove overlapping neighboring frames using homographic transformations based on special points is proposed. A method for comparing (and searching for coincidences) aerial imagery and similar satellite imagery that approximates data from different sources is also proposed. Given the result, the introduction of machine learning and artificial intelligence methods in geospatial data analysis can be the key to the development of highly automated systems that can adapt and respond to changes in large data streams.