The research in this doctoral thesis covers the area of exhaust gas emissions from marine diesel engines and the problem of estimating the amount of emissions. The quantification of emissions from ship powerplant is a challenge when creating an emissions inventory for maritime transport. One of the most important variables when calculating the total amount of emissions is the emission factor for certain exhaust gases. Emission factors depend on parameters such as the type of used fuel, the characteristics of the main and auxiliary engines, the engine load and the quality of the combustion process. In the thesis, research was carried out on three Ro-Pax ships sailing in the area of the Adriatic Sea between the ports of Croatia, Italy and Montenegro. The research includes measurements of the concentration of exhaust gases on ships in real-time exploitation conditions during various ship activities. Measurements were conducted with a Testo 350 Maritime device for measuring the content of exhaust gases, and a CO2 and NOx database was collected during the cruising and maneuvering phases. The measurement results represent exhaust gas emissions from the main engines depending on the engine load and the ship's activity. After the measurement and database collection process, the data was processed with the machine learning method using Gauss nonlinear regression, where the CO2 and NOx variables depend on the load of the main engines, the exhaust gas temperature and the oxygen content in the exhaust gas, which represents three independent variables. The modeling results are shown in diagrams where the measured values are compared with the estimated values. The optimization of the model depends on the choice of the appropriate covariance, and the obtained CO2 and NOx models were optimized in the cruising and maneuvering phases, which was indicated with a high coefficient of determination. Furthermore, the proposed models enable the selection of emission factors depending on the pollutant, engine load and ship activity. The estimated emission factors were compared with the emission factors from the relevant emissions inventory studies, where their advantages and possibilities of use were described. The correlation between emission factors and variable exploitation conditions was proven in the thesis, and the proposed models enable more reliable CO2 and NOx emission factors than currently available. By using adequate emission factors, it is possible to improve the estimation of total emissions from ship powerplants, and additional possibilities for improving research on this topic are proposed.