Image Processing and Object Detection for Advanced Driver Assistance Systems

Nowadays, the people irresponsibility and incorrect behaviours are pointed out as the main cause of automobile accidents. The vision of autonomous driving promises huge impacts on modern society. Its concept aims to improve the quality of human life by preventing accidents, managing the traffic, improving the comfort and safety, and reducing polluting gases. In the last years, this area noticed an outstanding evolution. However, a full autonomous system has not been conceived yet. This project was designed to address the previous necessity by creating a perception module for advanced driver assistance systems. To develop this system, many tools were used, namely: real-world data from a dataset, a deep learning model, the robot operating system framework, and image and point cloud processing algorithms. The work included the data processing of a stereo vision system as well as the data processing of a LiDAR sensor. At last, the extracted information was fused to reinforce the obstacle detection, making the perception module more robust. The Image Processing and Object Detection for Advanced Driver Assistance Systems revealed some promising results which can encourage the development of future projects