Date of Award

4-2012

Embargo Period

2-19-2013

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Robotics Institute

Advisor(s)

Chris Urmson, David Wettergreen

Abstract

Maps are important for both human and robot navigation. Given a route, drivingassistance systems consult maps to guide human drivers to their destinations. Similarly, topological maps of a road network provide a robotic vehicle with information about where it can drive and what driving behaviors it should use. By providing the necessary information about the driving environment, maps simplify both manual and autonomous driving.

The majority of existing cartographic databases are built, using manual surveys and operator interactions, to primarily assist human navigation. Hence, the resolution of existing maps is insufficient for use in robotics applications. Also, the coverage of these maps fails to extend to places where robotics applications require detailed geometric information.

To augment the resolution and coverage of existing maps, this thesis investigates computer vision algorithms to automatically build lane-level detailed maps of highways and parking lots by analyzing publicly available cartographic resources, such as orthoimagery.

Our map-building methods recognize image patterns and objects that are tightly coupled with the structure of the underlying road network by 1) identifying, without human intervention, locally consistent image cues and 2) linking them based on the obtained local evidence and prior information about roadways. We demonstrate the accuracy of our bootstrapping approach in building lane-level detailed roadwaymaps through experiments.

Due to expected abnormal events on highways such as roadwork, the geometry and traffic rules of highways that appear on maps can occasionally change. This thesis also addresses the problem of updating the resulting maps with temporary changes by analyzing perspective imagery acquired from a vision sensor installed on a vehicle.

To robustly recognize highway work zones, our sign recognizer focuses on handling variations of signs’ colors and shapes. Sign recognition errors, which are inevitable, can cause our system to misread temporary highway changes. To handle potential errors, our method utilizes the temporal redundancy of sign occurrences and their corresponding classification decisions. We demonstrate the effectiveness and robustness of our approach highway workzone recognition through testing with video data recorded under various weather conditions.

Two major results of this thesis work are 1) algorithms that analyze orthoimages to produce lane-level detailed maps of highways and parking lots and 2) on-vehicle computer vision algorithms that are able to recognize temporary changes on highways. Our maps can provide detailed information about a route, in advance, to either a human driver or a self-driving vehicle. While driving on highways, our roadway-assessing algorithms enable the vehicle to update the resulting maps with temporary changes to the route.

Comments

CMU-RI-TR-12-13

Share

COinS