Details

В данной работе, была изучена специфика компьютерного зрения в беспилотных транспортных средствах, приведены алгоритмы и категории машинного обучения, перечислены недостатки и пути решения связанные с компьютерным зрением, рассмотрены модели по распознаванию дорожных знаков и разметки на основе видеоизображений разных форматов. В ходе работы была протестирована работоспособность моделей в среде MATLAB, таких как:  1) Lane Departure Warning System  2) Color-based Road Tracking  3) Automate Ground Truth Labeling for Semantic Segmentation  4) Automate Ground Truth Labeling of Lane Boundaries  5) Traffic Warning Sign Recognition  После практической проверки корректности моделей был проведен анализ полученных данных и предложены рекомендации по области их применения.

In this article, the specifics of computer vision in self-driving cars were studied, algorithms and categories of machine learning were given, the disadvantages of computer vision and ways to solve them were listed, classical detection methods were considered and recognition systems for road signs and markings based on video images of various formats obtained from a car DVR were analyzed. The work tested the operability of models in the MATLAB environment, such as: 1) Lane Departure Warning System 2) Color-based Road Tracking 3) Automate Ground Truth Labeling for Semantic Segmentation 4) Automate Ground Truth Labeling of Lane Boundaries 5) Traffic Warning Sign Recognition After practical verification of the correctness of the models, the analysis of the data obtained was carried out and recommendations on the scope of their application were proposed.