Research of models and methods for practical implementation of unmanned transport systems for marine passenger terminals

Abstract

Two scenarios for the use of unmanned systems for marine passenger terminals that are identification of passengers at the terminal and tracking of moving objects are considered. The experimental study and the developed automation routines for identifying passengers were carried out on the basis of the laboratory of unmanned aircraft systems of Saint-Petersburg State University of Aerospace Instrumentation. For passenger terminal models, flight tasks of unmanned systems are simulated in the specialized Gazebo environment. Based on a series of experiments, it is found that the chosen method of histogram of oriented gradients provides a high level of accuracy and reliability. It is noted that the Matthews correlation coefficient, which reaches 95.09%, indicates a high degree of consistency and quality of binary classification of machine learning methods. During testing, the oriented gradient histogram method has showed that those with lower results in accuracy have received high results in sensitivity (reminder), which is 97.53%. This indicates that the use of this method can effectively minimize the number of false negative results, which is especially important in tasks where missing and losing an object during its identification can have serious consequences. As a result of the research carried out, the effectiveness of using the presented method is proven. The effectiveness of the developed new subprograms for passengers and other specified objects automatic identification, as well as the identification and tracking of objects for the first time when they are considered in the systems of marine passenger terminals, is also proven. It facilitates management and contributes to increased safety levels and monitoring effectiveness.