Maritime Object Detection, Tracking, and Classification Using Lidar and Vision-Based Sensor Fusion

<p>Autonomous Surface Vehicles have the capability of replacing dull, dirty, and dangerous jobs in the maritime field. However, few successful ASV systems exist today, as there is a need for greater sensing capabilities. Furthermore, a successful ASV system requires object detection and recognition capabilities to enable autonomous navigation and situational awareness. This thesis demonstrates an application of LiDAR sensors in maritime environments for object detection, classification, and camera sensor fusion. This is accomplished through the integration of a high-fidelity GPS/INS system, 3D LiDAR sensors, and a pair of cameras. After rotating LiDAR returns into a global reference frame, they are reduced to a 3D occupancy grid. Objects are then extracted and classified with a Support Vector Machine (SVM) classifier. The LiDAR returns, when converted from a global frame to a camera frame, then allow the cameras to process a region of their imaging frame to assist in the classification of objects using color-based features. The SVM implementation results in an overall accuracy 98.7% for 6 classes. The transformation into pixel coordinates is shown here to be successful, with an angular error of 2 degrees, attributed to measurement error propagated through rotations.</p>

Name thesis:degree_name

Master of Science in Mechanical Engineering

Level thesis:degree_level

Thesis - Open Access

Discipline thesis:degree_discipline

Mechanical Engineering

Year

2017

Author dc:creator

• Thompson, David John

• maritime
• object detection
• lidar
• vision-based
• Mechanical Engineering
• Oceanography and Atmospheric Sciences and Meteorology

Repository record dc:identifier

https://commons.erau.edu/edt/377

OAI identifier oai:identifier

oai:commons.erau.edu:edt-1376