Modeling an Image Clustering Algorithm For Detecting Overheated Railway Axle

This paper presents an approach to image identification based on a clustering algorithm for detecting the thermal characteristics of railway axle boxes. The study analyzes the potential of clustering algorithms in the task of digital image identification. The principles of the proposed algorithm are described in the context of image segmentation and compression, as well as pattern recognition in the transportation domain. A block diagram of the algorithm is provided along with an explanation of its operational principles. A scheme for implementing a machine vision system using the proposed algorithm is suggested for the detection of overheated axle boxes in railway transport. Experimental modeling was conducted for image segmentation based on color models in the infrared spectrum to identify regions with elevated temperatures in the MATLAB environment. For the experiment, thermal images of rolling stock axle boxes were selected one representing a normal condition and the other representing an overheated condition. The simulation results demonstrated that segmentation based on image color models allows for accurate delineation of the thermal characteristics of axle box images. The study confirmed that the use of the proposed algorithm and its software implementation for thermographic cameras is effective for recognizing temperature indicators. Additionally, the algorithm enables image compression, which increases data processing speed and facilitates the monitoring of thermal characteristics of rolling stock at high speeds.