LiDAR Pedestrian Detector and Semi-Automatic Annotation Tool for Labeling of 3D Data
(2019) In Master's Theses in Mathematical Sciences FMAM05 20191Mathematics (Faculty of Engineering)
- Abstract
- The goal of this Master's Thesis is to successfully detect and classify humans in a LiDAR data stream. The focus of the Thesis is on the detection and classification, not on the LiDAR technology.
To classify humans machine learning was used and to train the machine learning model we collected our own data and annotated it. A custom software was made for speeding up the annotation process.
The process for detecting humans in a scene was to first sweep the scene with a fixed size box which contain a point cloud. These point clouds were then split up by a clustering algorithm. Finally features were extracted from the clusters and classified using a classification algorithm.
The algorithm of choice for prediction became the Random... (More) - The goal of this Master's Thesis is to successfully detect and classify humans in a LiDAR data stream. The focus of the Thesis is on the detection and classification, not on the LiDAR technology.
To classify humans machine learning was used and to train the machine learning model we collected our own data and annotated it. A custom software was made for speeding up the annotation process.
The process for detecting humans in a scene was to first sweep the scene with a fixed size box which contain a point cloud. These point clouds were then split up by a clustering algorithm. Finally features were extracted from the clusters and classified using a classification algorithm.
The algorithm of choice for prediction became the Random Forest classifier which successfully classified unobstructed humans in different environments but occasionally gave false positives. (Less) - Popular Abstract
- LiDAR has been a fast growing technology recent years, mainly because of its use in autonomous vehicles and for mapping of terrains. In this project a pedestrian detector was created using a LiDAR and machine learning.
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8993073
- author
- Andersson, Roy LU and Andersson, Erik LU
- supervisor
- organization
- course
- FMAM05 20191
- year
- 2019
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- LiDAR, Machine Learning, Pedestrian Detection, Annotation, DBSCAN, SVM
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUTFMA-3390-2019
- ISSN
- 1404-6342
- other publication id
- 2019:E54
- language
- English
- additional info
- Last edited 27/08/2019
- id
- 8993073
- date added to LUP
- 2019-09-05 14:00:42
- date last changed
- 2019-09-05 14:00:42
@misc{8993073, abstract = {{The goal of this Master's Thesis is to successfully detect and classify humans in a LiDAR data stream. The focus of the Thesis is on the detection and classification, not on the LiDAR technology. To classify humans machine learning was used and to train the machine learning model we collected our own data and annotated it. A custom software was made for speeding up the annotation process. The process for detecting humans in a scene was to first sweep the scene with a fixed size box which contain a point cloud. These point clouds were then split up by a clustering algorithm. Finally features were extracted from the clusters and classified using a classification algorithm. The algorithm of choice for prediction became the Random Forest classifier which successfully classified unobstructed humans in different environments but occasionally gave false positives.}}, author = {{Andersson, Roy and Andersson, Erik}}, issn = {{1404-6342}}, language = {{eng}}, note = {{Student Paper}}, series = {{Master's Theses in Mathematical Sciences}}, title = {{LiDAR Pedestrian Detector and Semi-Automatic Annotation Tool for Labeling of 3D Data}}, year = {{2019}}, }