Petrovskyi M. Methods and hardware and software for specialized computer systems for image classification spots of laser beams

The object of research is the process of image processing of spots of laser beams. The aim is to improve the efficiency and speed of image processing of spots of laser beams through the creation methods and software for hardware specialized computer systems based on FPGA. Research methods include digital signal processing techniques; methods of mathematical analysis and mathematical modeling; theory of computer systems and networks; methods for creating hardware and software systems; techniques of high-level programming languages and hardware description languages; methods of measurement and metrology. The proposed method of section of image spots of laser beams, a feature of which is the use of the shape factor, which allows to compare them for belong to a class and determine their coordinates in real time; for the first time proposed method of image processing spots of laser beams, a feature of which is the combination of the method of sections and the method of direct parallel-hierarchical transformation that allows classify images and improving forecasting of their coordinate of energy centers with high accuracy by 5-10%; for the first time proposed choice of characteristics of the operator of recurrent convertion of data vector for the parallel-hierarchical networks and offer it basic structure for SIMD and MISD architecture, thus reducing power consumption in the calculation compared to the implementation on a PC; further development received method of direct parallel-hierarchical transformation, which differs from the known construction because use the improved architecture of a computer system which combining operators shift and transposition that reduces the number of memory accesses in 2 times and form a vector of parallel-hierarchical transformation for one clock cycle. The practical results consist in developing methods and tools that provide an opportunity to get hardware and software of fast image classification and their parallel implementation that allows image processing in real time based on FPGA; and moreover the resulting are basic blocks for creation parallel-hierarchical networks based on FPGA for processing and classification of images, that allowing to create parallel-hierarchical networks necessary scale and speed. The degree of implementation - the results obtained in the thesis were implemented: in central communication Station Ukrzaliznytsi within the research work "Development of algorithms of the automated system to support maintenance of transport communications network"; JSC "KIA Systems", during the modernization of the system of information transmission; in the learning process of the State Economic and Technological University of Transport. The scope (area) of using - communication, navigation, image processing.