Development of a hardware-software solution for detection of complex-shaped objects in video stream

The object of the study is the process of diagnosing complex-shaped objects in a video stream. The paper investigates the applied problem of creating a hardware-software solution for detecting complex-shaped objects in a video stream. Single-board computers Raspberry Pi models 4 and 5 with additional UPS HAT (D) modules and 21700 batteries were used as hardware, ensuring operation in the absence of power supply. Serial Camera Interface cameras and Full HD 1080p webcams were connected to the single-board computers to study effective methods of video processing using several studied video processing architectures. Eight video processing architectures based on the Oriented Features from Accelerated Segment Test and Rotated Binary Robust Independent Elementary Features and Scale-Invariant Feature Transform methods were considered. Each video processing architecture was tested using a one-minute video, where its average performance was determined. The limitations of video processing were a region of interest of 400x300 pixels and the presence of a limited number of reference images. To automate the launch of programs on single-board computers, the systemd initialization system was used. Known video processing algorithms were considered and a modification of the algorithm was proposed by using a double check for the presence of an object in the video stream. A hardware-software solution was implemented, consisting of a single-board computer with external cameras connected to it, and software for detecting complex-shaped objects in the video stream was created. The solution is useful as an auxiliary tool for detecting complex-shaped objects in the video stream on robotic platforms, in industry, everyday life, the educational process, and when repairing electronic modules. The practical significance of the study lies in the fact that the architecture for processing complex-shaped objects has been further developed. They provide for a double check for the presence of an object in the video stream, which increases the processing time of one frame, and on the other hand, increases the efficiency of object detection based on only one reference photo.