30 hp - Master's Thesis: Object Detection LiDARs Images

Are you ready to make a significant impact on the future of autonomous vehicles? Join us on a journey to advance the automotive industry's cutting-edge technology through your Master's thesis project on Object detection in LiDARs images! 

Background 

As the automotive industry rapidly evolves towards autonomous driving, ensuring the safety of self-driving vehicles is paramount. LiDARs are a crucial sensing technology for providing autonomous vehicles with information about the surrounding 3D environment. Various methods have been proposed in the literature for processing LiDAR data. Some of them are computationally expensive, such as processing the unstructured point cloud. The most common approach is to quantize the 3D cloud to a voxel grid and/or project it to a Bird-Eye-View (BEV) grid. A less commonly used approach is to process the data as an image. Many modern LiDARs are in fact delivering 3D data structured as images, in which depth is one of the channels.  

Scope 

The scope of this project is to develop Deep Neural Network object detection based on LiDAR images. The focus is on computational efficiency and compatibility with Scania’s Autonomous Vehicles Platform. The project will pay particular attention to the capabilities of using LiDAR images for detecting distant objects when both the ego vehicle and the other road used are moving at high speeds.  

Tasks 

The work will include: 

• Getting familiar with the literature and methods relevant for 3D object detection in lidar images, 

• Evaluating the performance of state-of-art solutions when applied to sensor data from heavy vehicles and benchmarking with other methods such as BEV, 

• Sensitivity analysis with respect to class, distance to the object, degree of occlusion, sensor resolution, weather/illumination conditions, 

• Investigating the use of transfer learning and/or multi-sensor fusion for improving detection performance. 

This thesis project offers a unique opportunity to delve deep into the world of LiDAR technology, machine learning, and autonomous vehicles. Join us in pushing the boundaries of automotive technology, improving safety, and paving the way for the future of autonomous driving. 

Applicant Requirements 

• A strong academic background in computer science, electrical engineering, robotics, or related fields. 

• Skills in analyzing and designing ML models for object detection and semantic segmentation. 

• Strong programming skills (e.g., Python, C++) and a background in computer vision or robotics are advantageous. 

• A passion for autonomous vehicles and sensor fusion technology. 

• Enthusiasm for research and the drive to tackle complex challenges in autonomous vehicle technology. 

• Experience with LiDAR systems, computer vision, and sensor fusion is a plus. 

Education/line/direction 

Number of students: 1 

Start date for the Thesis project: winter 2025 

Estimated timescale: 5 months 

Supervisors 

Bogdan Timus, EEARP, tbogdan.timus@scania.com

Alexandre Miro, EEARP, alexandre.justo.miro@scania.com

Application 

Your application should contain CV, motivation letter and copies of grades 

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.     

References 

[1] Bichen Wu et al., "SqueezeSeg: Convolutional Neural Nets with Recurrent CRF for Real-Time Road-Object Segmentation from 3D LiDAR Point Cloud", 2017 

[2] Meyer, Gregory, “LaserNet: An efficient probabilistic 3d object detector for autonomous driving, 2019 

[3] Liang, Zhiong; “RangeRCNN: Towards fast and accurate 3d object detection with range image representation”, 2020 

[4] Fan, Lue, “Rangedet: In defense of range view for lidar-based 3d object detection”, 2021 

Images", 2022 

[6] Benjamin Wilson et al., "What Matters in Range View 3D Object Detection" 2024