Algorithm for marking related areas in streaming image processing

Abstract

Algorithm for marking related areas in streaming image processing

Kovalev A.V., Bakumenko A.N.

Incoming article date: 13.03.2022

The problem of choosing the desired image segmentation algorithm and its real-time implementation arises in systems that work with a high-speed video data stream. Labeling related areas is an important step in preparing images for neural network training set. In this paper a modified one-pass image labeling algorithm for use in field-programmable gate arrays (FPGAs), as well as its implementation using the Xilinx System Generator for DSP and the Matlab/Simulink package was proposed. The results of the algorithm operation on test images, as well as images obtained experimentally, are presented. This algorithm was developed to analyze and evaluate the quality of wheat seeds by classifying their images using artificial intelligence.

Keywords: labeling of related areas, artificial intelligence, segmentation, stream processing