SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
This paper presents an in-depth performance analysis and validation of a correlation based stereo vision system being used as part of the ongoing 2003 Mars Exploration Rover flight mission.
A Mars rover must be able to sense its local environment with sufficient resolution and accuracy to avoid local obstacles and hazards while moving a significant distance each day. Power efficiency and reliability are extremely important considerations, making stereo correlation an attractive method of range sensing compared to laser scanning, if the computational load and correspondence errors can be handled. Techniques for treatment of these problems, including the use of more than two cameras to reduce correspondence errors and possibly to limit the computational burden of stereo processing, have been tested at JPL. Once a reliable range map is obtained, it must be transformed to a plan view and compared to a stored terrain database, in order to refine the estimated position of the rover and to improve the database. The slope and roughness of each terrain region are computed, which form the basis for a traversability map allowing local path planning. Ongoing research and field testing of such a system is described.
A stereo correlation method on the object domain is proposed to generate the accurate and dense Digital Elevation Models (DEMs) from lunar orbital imagery. The NASA Ames Intelligent Robotics Group (IRG) aims to produce high-quality terrain reconstructions of the Moon from Apollo Metric Camera (AMC) data. In particular, IRG makes use of a stereo vision process, the Ames Stereo Pipeline (ASP), to automatically generate DEMs from consecutive AMC image pairs. Given camera parameters of an image pair from bundle adjustment in ASP, a correlation window is defined on the terrain with the predefined surface normal of a post rather than image domain. The squared error of back-projected images on the local terrain is minimized with respect to the post elevation. This single dimensional optimization is solved efficiently and improves the accuracy of the elevation estimate.
The camera registration extracted from feature based stereo is usually considered sufficient to accurately localize the 3D points. However, for natural scenes the feature localization is not as precise as in man-made environments. This results in small camera registration errors. We show that even very small registration errors result in large errors in dense surface reconstruction. We describe a method for registering entire images to the inaccurate surface model. This gives small, but crucially important improvements to the camera parameters. The new registration gives dramatically better dense surface reconstruction.
This paper introduces an advanced rover localization system suitable for autonomous planetary exploration in the absence of Global Positioning System (GPS) infrastructure. Given an existing terrain map (image and elevation) obtained from satellite imagery and the images provided by the rover stereo camera system, the proposed method determines the best rover location through visual odometry, 3D terrain and hori- zon matching. The system is tested on data retrieved from a 3 km traverse of the Basalt Hills quarry in California where the GPS track is used as ground truth. Experimental results show the system presented here reduces by over 60 the localization error obtained by wheel odometry.
Parallel relaxation computations such as those of connectionist networks offer a useful model for constraint integration and intrinsic image combination in developing a general-purpose stereo matching algorithm. This paper describes such a stereo algorithm that incorporates hierarchical, surface-structure, and edge-appearance constraints that are redefined and integrated at the level of individual candidate matches. The algorithm produces a high percentage of correct decisions on a wide variety of stereo pairs. Its few errors arise when the correlation measures defined by the constraints are either weakened or ambiguous, as in the case of periodic patterns in the images. Two additional mechanisms are discussed for overcoming the remaining errors.
Digital terrain models produced by computer correlation of stereo images are likely to contain occasional gross errors in terrain elevation. These errors typically result from having mismatched sub-areas of the two images, a problem which can occur for a variety of image- and terrain-related reasons. Such elevation errors produce undesirable effects when the models are further processed, and should be detected and corrected as early in the processing as possible. Algorithms have been developed to detect and correct errors in digital terrain models. These algorithms focus on the use of constraints on both the allowable slope and the allowable change in slope in local areas around each point. Relaxation-like techniques are employed in the iteration of the detection and correction phases to obtain best results.
An important phase of any Stereo Imaging Velocimetry experiment is particle tracking. Particle tracking seeks to identify and characterize the motion of individual particles entrained in a fluid or air experiment. We analyze a cylindrical chamber filled with water and seeded with density-matched particles. In every four-frame sequence, we identify a particle track by assigning a unique track label for each camera image. The conventional approach to particle tracking is to use an exhaustive tree-search method utilizing greedy algorithms to reduce search times. However, these types of algorithms are not optimal due to a cascade effect of incorrect decisions upon adjacent tracks. We examine the use of a guided evolutionary neural net with simulated annealing to arrive at a globally optimal assignment of tracks. The net is guided both by the minimization of the search space through the use of prior limiting assumptions about valid tracks and by a strategy which seeks to avoid high-energy intermediate states which can trap the net in a local minimum. A stochastic search algorithm is used in place of back-propagation of error to further reduce the chance of being trapped in an energy well. Global optimization is achieved by minimizing an objective function, which includes both track smoothness and particle-image utilization parameters. In this paper we describe our model and present our experimental results. We compare our results with a nonoptimizing, predictive tracker and obtain an average increase in valid track yield of 27 percent
Onboard localization capabilities for planetary rovers to date have used relative navigation, by integrating combinations of wheel odometry, visual odometry, and inertial measurements during each drive to track position relative to the start of each drive. At the end of each drive, a “ground-in-the-loop” (GITL) interaction is used to get a position update from human operators in a more global reference frame, such as a map frame defined by orbital reconnaissance imaging of a large region around the rover’s current position. For Mars rovers, this typically has involved downlinking imagery from the rover mast cameras and using interactive visualization tools on Earth to register such images to the orbital reconnaissance images. For safety purposes, rover mission operations typically specify “keep out zones”, which human operators recognize as being unsafe in the orbital images. Autonomous rover drives are limited in distance so that accumulated relative navigation error does not risk the possibility of the rover driving into a keep out zone. The allowable autonomous drive distance in this mode of operation depends on the distribution of keep out zones and the accuracy of relative navigation; in practice, drive limits of a few hundred meters between GITL cycles are to be expected. Several rover mission concepts have recently been studied that require much longer drives between GITL cycles, particularly for the Moon. This includes lunar rover mission concepts that involve (1) driving mostly in sunlight at low latitudes, (2) driving in permanently shadowed regions near the south pole, and (3) a mixture of day and night driving in mid-latitudes. These concepts include total traverse distance requirements of up to 1,800 km in 4 Earth years, with individual drives of several kilometers between stops for downlink. These concepts require greater autonomy to minimize GITL cycles to enable such large range; onboard global localization is a key element of such autonomy. Multiple techniques have been studied in the past for onboard rover global localization, including radio navigation aiding from an orbiter, recognizing horizon landmarks that are known in a regional elevation map, and correlating a local elevation map created onboard the rover with a regional elevation map. These techniques all have drawbacks, including requiring an expensive extra mission element (navigation orbiter), unavailability of sufficient regional elevation map data, or limited accuracy in resulting position estimates (e.g. a few hundred meters with horizon landmarks). For the Moon, the ubiquitous craters offer another possibility, which involves mapping craters from orbit, then recognizing crater landmarks with cameras and/or a lidar onboard the rover. This approach is applicable everywhere on the Moon, does not require high resolution stereo imaging from orbit as some other approaches do, and has potential to enable position knowledge with order of 10 m accuracy at all times. This paper will provide more detail on our technical approach to crater-based lunar rover localization and will present initial results on crater detection using 3-D point cloud data from onboard lidar or stereo cameras and using shading cues in monocular onboard imagery.
Virtual acoustics, also known as 3-D sound and auralization, is the simulation of the complex acoustic field experienced by a listener within an environment. Going beyond the simple intensity panning of normal stereo techniques, the goal is to process sounds so that they appear to come from particular locations in three-dimensional space. Although loudspeaker systems are being developed, most of the recent work focuses on using headphones for playback and is the outgrowth of earlier analog techniques. For example, in binaural recording, the sound of an orchestra playing classical music is recorded through small mics in the two "ear canals" of an anthropomorphic artificial or "dummy" head placed in the audience of a concert hall. When the recorded piece is played back over headphones, the listener passively experiences the illusion of hearing the violins on the left and the cellos on the right, along with all the associated echoes, resonances, and ambience of the original environment. Current techniques use digital signal processing to synthesize the acoustical properties that people use to localize a sound source in space. Thus, they provide the flexibility of a kind of digital dummy head, allowing a more active experience in which a listener can both design and move around or interact with a simulated acoustic environment in real time. Such simulations are being developed for a variety of application areas including architectural acoustics, advanced human-computer interfaces, telepresence and virtual reality, navigation aids for the visually-impaired, and as a test bed for psychoacoustical investigations of complex spatial cues. The tutorial will review the basic psychoacoustical cues that determine human sound localization and the techniques used to measure these cues as Head-Related Transfer Functions (HRTFs) for the purpose of synthesizing virtual acoustic environments. The only conclusive test of the adequacy of such simulations is an operational one in which the localization of real and synthesized stimuli are directly compared in psychophysical studies. To this end, the results of psychophysical experiments examining the perceptual validity of the synthesis technique will be reviewed and factors that can enhance perceptual accuracy and realism will be discussed. Of particular interest is the relationship between individual differences in HRTFs and in behavior, the role of reverberant cues in reducing the perceptual errors observed with virtual sound sources, and the importance of developing perceptually valid methods of simplifying the synthesis technique. Recent attempts to implement the synthesis technique in real time systems will also be discussed and an attempt made to interpret their quoted system specifications in terms of perceptual performance. Finally, some critical research and technology development issues for the future will be outlined.