Public defence in Interactive Systems, M.Sc. Aziza Zhanabatyrova

The title of the thesis: Crowdsourced 3D semantic mapping and change detection in urban driving environments

Thesis defender: Aziza Zhanabatyrova
Opponent: Dr. Fabio Remondino, Bruno Kessler Foundation, Italy
Custos: Prof. Yu Xiao, Aalto University School of Electrical Engineering

The study focuses on creating cost-effective and scalable solutions to build and update accurate semantic maps for autonomous vehicles. The purpose of the research is to address the challenges of dynamic urban environments by detecting and localizing changes, such as new traffic signs, using crowdsourced data from consumer-grade cameras (e.g., smartphones, dashboard cameras) and traffic flow information from existing mapping services. 

This work is highly relevant to ongoing advancements in autonomous driving, as traditional approaches rely on specialized vehicles equipped with expensive sensors, making them time-consuming and costly to operate. The research introduces a novel pipeline that uses Structure from Motion (SfM) to generate 3D point clouds from visual data, enabling precise change detection and localization. Additionally, it proposes a deep-learning approach for coarse-grained change detection using traffic flow data, narrowing down areas for refinement. 

Key results include a comprehensive comparison of SfM techniques on crowdsourced visual data and guidelines for constructing large-scale 3D maps to maintain accuracy and reliability. This research contributes significant new insights into leveraging visual crowdsourcing and traffic data for semantic mapping. 

The findings have practical applications in deploying cost-effective automatic change detection systems, allowing real-time updates of semantic maps. This enhances autonomous vehicle navigation and safety while minimizing the need for expensive hardware. 

In conclusion, the research lays the groundwork for scalable, accurate, and cost-efficient techniques for semantic mapping and maintenance, enabling better integration of autonomous vehicles into urban environments.

Key words: crowdsourcing, structure from motion, change detection, 3d mapping, autonomous driving, semantic maps

Thesis available for public display 10 days prior to the defence at Aaltodoc. 

Contact:
Linkedin http://linkedin.com/in/azizazhanabatyrova 
Phone +358504718328 
email zhanabatyrova.aziza@aalto.fi