The thesis was done in cooperation with the HAMK Tech research unit located in Hämeenlinna. HAMK Tech is a research unit of Häme University of Applied Sciences, whose research areas are materials for manufacturing industry, design, and manufacturing technologies, construction, and the energy efficiency of built environment. The HAMK Tech research unit has several projects each year that investigate the soiling of various materials. The goal was to investigate how machine vision could be utilized in the analysis of these soiling samples. With the help of machine vision, it was hoped to find a way to analyze and calculate the amount of dirt in the samples more precisely, and to reduce the amount of manual work. The research object was limited to samples painted on thin metal sheets that had been soiled in various ways. The implementation phase of the thesis consisted of photographing soiling samples with different cameras, processing the images, implementing and programming a program that is suitable for image processing, running the samples with the program, and analyzing the results. Various digital cameras and a hyperspectral camera were used to photograph the soiling samples. The machine vision programs were chosen according to their features and usage possibilities to meet the research topic. An experiment was also conducted for the thesis, in which the participants ranked the samples of digital images analyzed by machine vision from the cleanest to the dirtiest in their own opinion. The results showed that the people who participated saw the pollution in a slightly different way than machine vision. People thought that the dirtiest samples were the ones where the amount of dirt is very visible and large. In the study, 25 samples were examined, and only for one sample did humans and machine vision agree on the degree of contamination. Spectra were analyzed from images that were taken with a hyperspectral camera and their correspondences to known impurities.