Master Thesis: “Lidar based pedestrian detection: aplication to the autonomous vehicle location”

Hi everyone, on December 20th, 2017 I finished my Master’s degree and I presented my Master Thesis “Lidar based pedestrian detection: aplication to the autonomous vehicle location”. I’d like to thank Jesús Manuel Gomez de Gabriel for his support and supervision.

Alberto presented his work on the use of LIDAR sensors to the detection of pedestrian from vehicles.

All of us have seen the amazing work that companies like Google, Tesla, Uber and Ford are doing to develop self-driving cars. But, ¿how do they work? Well, in my master thesis I focus in one specific task: pedestrian detection and tracking using a lidar sensor. This kind of sensor can measure distances from sensor to an object sending a pulsed laser beam and it’s very useful for mapping the environment surrounding the vehicle.

Lidar data using RPLidar

The 2D point cloud obtained through sensor is clustering in segments as a first step before people detection. Once fragments are splitted, we extract features from segments that allow us to differentiate a person from others obstacles in the scene. The next step is to label segments as people or obstacles in order to train a deep learning algorithm named Adaboost. This algorithm is a supervised machine learning that can separate a data set into two parts (person / no person).

Last, segments labeled as people are tracked to avoid a possible collision with the vehicle. This video shows the process.

I’d finally like to thank Jesús Manuel Gomez de Gabriel for his support and supervision in the development of this project.

Leave a Reply

Your email address will not be published. Required fields are marked *