NASA Technical Reports Server (NTRS) 20110003036: Visual Object Recognition and Tracking of Tools

0.13 MB·English

by  NASA Technical Reports Server (NTRS)

Information Sciences
Modeling Electromagnetic Scattering From Complex Inhomo-
geneous Objects
Complex, inhomogeneous objects can be easily modeled using commercial CAD packages.
NASA Langley Research Center, Hampton, Virginia
This software innovation is designed FEM, whereas the electromagnetic MOS/M, common boundaries where
to develop a mathematical formulation fields outside the fictitious boundary boundary conditions are to be imple-
to estimate the electromagnetic scatter- are obtained using the assumed equiva- mented are also identified. The data
ing characteristics of complex, inhomo- lent electric and magnetic currents file created by the COSMOS/M for
geneous objects using the finite-ele- flowing on the fictitious boundary. node and element information is then
ment-method (FEM) and Continuity of the electromagnetic generated and processed through the
method-of-moments (MoM) concepts, fields across the fictitious boundary re- preprocessor part of FEMOM3DS code
as well as to develop a FORTRAN code sults in partly sparse/dense matrix that to create edge information and node
called FEMOM3DS (Finite Element is solved for the unknown fields inside information. The preprocessed data are
Method and Method of Moments for 3- the fictitious boundary including the then run through the main part of
Dimensional Scattering), which will im- boundary surface. FEMOM3DS to obtain electromagnetic
plement the steps that are described in The steps laid out in the mathemati- scattering. The output files from the
the mathematical formulation. Very cal formulation are carried out in FEMOM3DS can be used for displaying
complex objects can be easily modeled, FEMOM3DS. Along with the main the results in a graphical format.
and the operator of the code is not re- FEMOM3DS, the innovation uses a The FEMOM3DS is written in FOR-
quired to know the details of electro- commercial computer-aided-design TRAN 77. The code was successfully
magnetic theory to study electromag- (CAD) package for geometrical model- compiled on a CONVEX machine.
netic scattering. ing, and a post-processing package such However, the code can be compiled on
Consider a complex inhomogeneous as Techplot for displaying graphically any 32-bit machine like PCs or SUN,
object geometry, the electromagnetic the results obtained using FEMOM3DS. SGI UNIX Station. To get correct re-
scattering characteristics of which are In the present FEMOM3DS code the sults, dimensions must be given in cen-
to be estimated when illuminated by a COSMOS/M commercial software is timeters, frequency of operation must
plane electromagnetic wave. To facili- used as a CAD tool to model the geom- be given in GHz and incident and ob-
tate the mathematical formulation etry of a given problem. The COS- servation angles must be specified in
using the hybrid FEM/MoM, the object MOS/M is also used to discretize the degrees.
is assumed to be enclosed by a surface FEM region using the tetrahedral ele- This work was done by Manohar Desh-
indicated as a fictitious outer bound- ments. Various inhomogeneous regions pande and C. J. Reddy of Langley Research
ary. Inside the boundary, the electro- are taken care of by having many parts Center. Further information is contained in a
magnetic fields are obtained using the in the FEM region. Using the COS- TSP (see page 1).LAR-17090-1
Visual Object Recognition and Tracking of Tools
This method can be used to track tools held and used by humans, such as surgical tools.
Lyndon B. Johnson Space Center, Houston, Texas
A method has been created to auto- and (3) refined type, pose, and geometry proach is used for condensing synthetic
matically build an algorithm off-line, (RTPG) calculation. The image segmen- images using an image subspace that
using computer-aided design (CAD) tation module files all the tools in an captures key information. Three degrees
models, and to apply this at runtime. image and isolates them from the back- of orientation, three degrees of position,
The object type is discriminated, and ground. For this, the system uses edge-de- and any number of degrees of freedom
the position and orientation are identi- tection and thresholding to find the pix- in geometry change are included. To do
fied. This system can work with a single els that are part of a tool. After the pixels this, a template-matching framework is
image and can provide improved per- are identified, nearby pixels are grouped applied. This framework uses an off-line
formance using multiple images pro- into blobs. These blobs represent the po- system for calculating template images,
vided from videos. tential tools in the image and are the measurement images, and the measure-
The spatial processing unit uses three product of the segmentation algorithm. ments of the template images. These re-
stages: (1) segmentation; (2) initial type, The second module uses matched fil- sults are used online to match seg-
pose, and geometry (ITPG) estimation; tering (or template matching). This ap- mented tools against the templates.
NASA Tech Briefs, January 2011 35
The final module is the RTPG proces- as a metric on the difference between the Further information is contained in a TSP
sor. Its role is to find the exact states of captured image and a synthetic image (see page 1).
the tools given initial conditions provided with an object in the given orientation, In accordance with Public Law 96-517,
by the ITPG module. The requirement distance, and articulation. The synthetic the contractor has elected to retain title to this
that the initial conditions exist allows this image is created using a model that is invention. Inquiries concerning rights for its
module to make use of a local search looked up in an object-model database. commercial use should be addressed to:
(whereas the ITPG module had global A composable software architecture is Energid Technologies
scope). To perform the local search, 3D used for implementation. Video is first 124 Mount Auburn Street
model matching is used, where a syn- preprocessed to remove sensor anom- Suite 200 North
thetic image of the object is created and alies (like dead pixels), and then is Cambridge, MA 02138
compared to the sensed data. The avail- processed sequentially by a prioritized Phone No.: [phone]
ability of low-cost PC graphics hardware list of tracker-identifiers. Toll Free No.: [phone]
allows rapid creation of synthetic images. This work was done by James English, Refer to MSC-23947-1, volume and num-
In this approach, a function of orienta- Chu-Yin Chang, and Neil Tardella of En- ber of this NASA Tech Briefsissue, and the
tion, distance, and articulation is defined ergid Technologies for Johnson Space Center. page number.
Method for Implementing Optical Phase Adjustment
Goddard Space Flight Center, Greenbelt, Maryland
A method has been developed to me- pendicular to that of the pump beam. the pump beam, its direction can be var-
chanically implement the optical phase The probe beam can be considered as ied relative to the plane defined by the
shift by adjusting the polarization of the defining the axis of a cylinder, while the probe electric field and the magnetic
pump and probe beams in an NMOR pump beam is directed radially. The field to be measured.
(nonlinear magneto-optical rotation) magnetic field to be measured defines a This work was done by David C. Hovde of
magnetometer as the proper phase shift third vector, but it is also taken to lie Southwest Sciences and Eric Corsini of the
is necessary to induce self-oscillation. along the cylinder axis. Both the pump University of California, Berkeley, for God-
This innovation consists of mounting and probe beams are polarized such dard Space Flight Center. For further informa-
the pump beam on a ring that sur- that their electric field vectors are sub- tion, contact the Goddard Innovative Part-
rounds the atomic vapor sample. The stantially perpendicular to the magnet nerships Office at (301) 286-5810.
propagation of the probe beam is per- field. By rotation of the ring supporting GSC-15608-1
Visual SLAM Using Variance Grid Maps
This algorithm is suitable for real-time navigation on irregular terrain.
NASA’s Jet Propulsion Laboratory, Pasadena, California
An algorithm denoted “Gamma- SLAM algorithms have been feature- jectories, and (2) finding the map con-
SLAM” performs further processing, in based and not suitable for real-time ap- ditioned on any given trajectory. The
real time, of preprocessed digitized im- plications and, hence, not suitable for factorization involves the use of a parti-
ages acquired by a stereoscopic pair of autonomous navigation on irregularly cle filter in which each particle encodes
electronic cameras aboard an off-road structured terrain. both a possible trajectory and a map
robotic ground vehicle to build accu- The Gamma-SLAM algorithm incor- conditioned on that trajectory. The base
rate maps of the terrain and determine porates two key innovations: estimate of the trajectory is derived
the location of the vehicle with respect (cid:129) Visual odometry (in contradistinction from visual odometry, and the map con-
to the maps. Part of the name of the al- to wheel odometry) is used to estimate ditioned on that trajectory is a Cartesian
gorithm reflects the fact that the the motion of the vehicle. grid of elevation variances. In compari-
process of building the maps and de- (cid:129) An elevation variance map (in contradis- son with traditional occupancy or eleva-
termining the location with respect to tinction to an occupancy or an elevation tion grid maps, the grid elevation vari-
them is denoted “simultaneous local- map) is used to represent the terrain. ance maps are much better for
ization and mapping” (SLAM). Most The Gamma-SLAM algorithm makes representing the structure of vegetated
prior real-time SLAM algorithms have use of a Rao-Blackwellized particle filter or rocky terrain.
been limited in applicability to (1) sys- (RBPF) from Bayesian estimation the- This work was done by Andrew B.
tems equipped with scanning laser ory for maintaining a distribution over Howard of Caltech and Tim K. Marks of the
range finders as the primary sensors in poses and maps. The core idea of the University of California San Diego for
(2) indoor environments (or relatively RBPF approach is that the SLAM prob- NASA’s Jet Propulsion Laboratory. For more
simply structured outdoor environ- lem can be factored into two parts: (1) information, contact [email protected].
ments). The few prior vision-based finding the distribution over robot tra- NPO-46114
36 NASA Tech Briefs, January 2011