Abstract. Target recognition (TR) is widely used in different military and civil applications and permits enhanced intelligence and autonomously operating platforms design. The article describes existing systems for TR such as deep learning aided computer vision; target tracking architecture, based on the tracking-by-detection paradigm; a target detection dataset; deep neural networks; a system for the management of a plurality of sensors; a target recognition architecture, adaptive to operational conditions and a target detection system, based on the theory of multi-temporal recognition. Unfortunately, the existing systems do not orient for real-time processing or can be applied for synthetic aperture radar images only, or used for image processing of soft targets, etc. This article presents the data regarding proposed new systems for targets recognition and determination of their parameters, based on central image chord transformation. The systems’ main processing units are described. The structures of the elaborated systems and the principles of their functioning are presented. The models of data processing flow in the systems are described. The determination of the processing time of the operations, realized in the systems was made and the estimation of the throughput of the systems was done. The optimization of the elaborated systems was made. The results regarding systems’ characteristics are presented.
Kew words: Target recognition, data processing flow, throughput, optimization.
Full Text (PDF): Download