2013 IR&D Annual Report

GPS-denied Localization System, 10-R8248

Principal Investigators
Kristopher C. Kozak
Christopher L. Lewis
Marc C. Alban
Samuel E. Slocum
Michael O. Blanton

Inclusive Dates:  09/06/11 – 09/06/13

Figure 1: Features of opportunity in ground images are matched by appearance subject to geometric constraints to identify overlapping frames and determine vehicle location.
Figure 1. Features of opportunity in ground images are matched by appearance subject to geometric constraints to identify overlapping frames and determine vehicle location.
Figure 2: Stereo imagery acquired on a ground vehicle can be used to determine location relative to a georegistered aerial image.
Figure 2. Stereo imagery acquired on a ground vehicle can be used to determine location relative to a georegistered aerial image.

Background — Global Positioning System (GPS) receivers provide a low-cost localization and navigation solution to a wide variety of commercial and military systems. As safety-critical systems come to rely on GPS (and other satellite-based localization systems), concerns have mounted due to its well-known vulnerabilities; GPS has limited accuracy and requires an unobstructed line-of-sight to multiple satellites. Its signals are subject to interference, multi-path, jamming and spoofing. While GPS has become more essential and ubiquitous, few practical alternatives have emerged. The fragility of GPS can be considered one of the limiting factors in the adoption of some cutting-edge technologies such as automated driving. The objective of this project was to develop a camera-based system that provides real-time localization measurements and can serve as a reliable supplement or alternative to GPS.

Approach — SwRI developed two related map-based localization methods that use cameras on a vehicle. Two camera systems were designed: a downward-facing camera with high intensity illumination and a forward-facing stereo pair. The downward-facing camera system allows for extremely high precision localization on pre-driven, mapped routes, while the forward-facing camera system allows for localization using both maps generated on pre-driven routes, as well as maps consisting of readily available aerial imagery. By processing the real-time imagery acquired with the cameras and comparing the image features/landmarks to the geo-referenced imagery that comprises the map, the location of the cameras (and thus the vehicle on which the cameras are mounted) can be determined. In addition, either camera system can be used to compute differential motion between camera frames, and thus can fill in gaps when the live imagery from the vehicle cannot be confidently matched to the map.

Accomplishments — For the ground-facing camera localization approach, a full-hardware system, which includes a camera and synchronized high-intensity illumination, was designed and installed on an SwRI vehicle. An algorithmic framework based on feature matching with geometric constraints was developed as the basis of the localization approach. This approach has been shown in a subsequent analysis to have a very high success rate for positively identifying location (on a variety of road surfaces), with a very low rate of incorrect matches. A map representation and optimization scheme was developed to facilitate live localization and to improve the overall consistency and accuracy of the localization. Finally, the full ground-facing system was implemented and operated successfully in real time on the vehicle. The precision of this system was measured versus a high-accuracy GPS, and found to perform comparably, on the order of 1 cm. For the forward-facing stereo camera localization approach, two separate algorithmic frameworks were developed, a feature matching approach and an image correlation approach. Both methods have been demonstrated to yield successful localization on several types of data, including real projected stereo imagery.

Benefiting government, industry and the public through innovative science and technology
Southwest Research Institute® (SwRI®), headquartered in San Antonio, Texas, is a multidisciplinary, independent, nonprofit, applied engineering and physical sciences research and development organization with 10 technical divisions.
04/15/14