In response to the challenges posed by SPAD (Signal Passed At Danger) events, a concept was developed for a system that informs the train driver of approaching light signals and does not require modifications to vehicle systems or a costly homologation process. The main objective of the work was to create a software layer for a system to recognize railway signal aspects in images from the front camera. For this purpose, image processing using cascade classifiers was applied, with the task of detecting light signals. The method is based on machine learning, using training images that contain the target objects as well as images without them. The developed tool was tested for correct detection of light signals and identification of their aspects, i.e. correct recognition of signal aspects, no recognition of a light signal, incorrect interpretation of an image or false positive detection of a color light signal. Test results showed that the designed system is able to detect railway color light signals in 95.6% of cases under good visibility conditions and correctly recognize color light signal aspects in 64.6% of cases. Tests conducted in various scenarios showed that the effectiveness of aspect detection is strongly dependent on external conditions such as sunlight intensity and atmospheric phenomena like fog or rain. The developed train driver assistance system, based on image processing, can provide a complementary support tool for rail vehicle drivers. There are plans for further work to improve the quality of visual recognition.