Carl F. Diegert

Geospatial image mining for nuclear proliferation detection: Challenges and new opportunities

With increasing understanding and availability of nuclear technologies, and increasing persuasion of nuclear technologies by several new countries, it is increasingly becoming important to monitor the nuclear proliferation activities. There is a great need for developing technologies to automatically or semi-automatically detect nuclear proliferation activities using remote sensing. Images acquired from earth observation satellites is an important source of information in detecting proliferation activities. High-resolution remote sensing images are highly useful in verifying the correctness, as well as completeness of any nuclear program. DOE national laboratories are interested in detecting nuclear proliferation by developing advanced geospatial image mining algorithms. In this paper we describe the current understanding of geospatial image mining techniques and enumerate key gaps and identify future research needs in the context of nuclear proliferation.

The emerging versatility of a scannerless range imager

Sandia National Laboratories is nearing the completion of the initial development of a unique type of range imaging sensor. This innovative imaging optical radar is based on an active flood-light scene illuminator and an image intensified CCD camera receiver. It is an all solid-state device (no moving parts) and offers significant size, performance, reliability, simplicity, and affordability advantages over other types of 3D sensor technologies, including: scanned laser radar, stereo vision, and structured lighting. The sensor is based on low cost, commercially available hardware, and is very well suited for affordable application to a wide variety of military and commercial uses, including: munition guidance, target recognition, robotic vision, automated inspection, driver enhanced vision, collision avoidance, site security and monitoring, terrain mapping, and facility surveying. This paper reviews the sensor technology and its development for the advanced conventional munition guidance application, and discusses a few of the many other emerging applications for this new innovative sensor technology.

Scannerless terrain mapper

NASA-Ames Research Center, in collaboration with Sandia National Laboratories, is developing a scannerless terrain mapper (STM) for autonomous vehicle guidance through the use of virtual reality. The STM sensor is based on an innovative imaging optical radar technology that is being developed by Sandia National Laboratories. The sensor uses active flood- light scene illumination and an image intensified CCD camera receiver to rapidly produce and record very high quality range imagery of observed scenes. The STM is an all solid- state device (containing no moving parts) and offers significant size, performance, reliability, simplicity, and affordability advantages over other types of 3-D sensor technologies, such as scanned laser radar, stereo vision, and structured lighting. The sensor is based on low cost, commercially available hardware, and is very well suited for affordable application to a wide variety of military and commercial uses, including: munition guidance, target recognition, robotic vision, automated inspection, driver enhanced vision, collision avoidance, site security and monitoring, and facility surveying. This paper reviews the sensor technology, discusses NASAs terrain mapping applications, and presents results from the initial testing of the sensor at NASAs planetary landscape simulator.

Practical automatic placement for standard-cell integrated circuits

SYNOPTIC ABSTRACTThis paper introduces the placement problem that arises in the established standard-cell-style design of integrated circuit (IC) parts, and reports on a method for effectively solving this problem. The placement problem is to assign geometric coordinates on the two-dimensional IC “chip” to its many small, predesigned logic functions (standard cells). Applying a practical, simulated-annealing search to the critical placement problem as part of designing an IC can have a surprisingly large impact on a parts cost and manufacturability.Results from a recent benchmarking contest illustrate the importance of effectively solving the placement problem, and support the effectiveness of the methods introduced in this paper. The new solution method applies simulated-annealing search using an approximate placement model (considering only bounding boxes of the logical nets, and considering only approximate, fixed-grid carrier locations for the cells that define the boxes), adaptive wiring congestion ...

A low-cost, high-resolution, video-rate imaging optical radar

Sandia National Laboratories has developed a unique type of portable low-cost range imaging optical radar (laser radar or LADAR). This innovative sensor is comprised of an active floodlight scene illuminator and an image intensified CCD camera receiver. It is a solid-state device (no moving parts) that offers significant size, performance, reliability, and simplicity advantages over other types of 3D imaging sensors. This unique flash LADAR is based on low- cost, commercially available hardware, and is well suited for many government and commercial uses. This paper presents an update of Sandias development of the Scannerless Range Imager technology and applications, and discusses the progress that has been made in evolving the sensor into a compact, low cost, high-resolution, video rate Laser Dynamic Range Imager.

Building accurate geometric models from abundant range imaging information

We define two simple metrics for accuracy of models built from range imaging information. We apply the metric to a model built from a recent range image taken at the laser radar Development and Evaluation Facility, Eglin AFB, using a scannerless range imager (SRI) from Sandia National Laboratories. We also present graphical displays of the residual information produced as a byproduct of this measurement, and discuss mechanisms that these data suggest for further improvement in the performance of this already impressive SRI.

A combinatorial method for tracing objects using semantics of their shape

We present a shape-first approach to finding automobiles and trucks in overhead images and include results from our analysis of an image from the Overhead Imaging Research Dataset [1]. For the OIRDS, our shape-first approach traces candidate vehicle outlines by exploiting knowledge about an overhead image of a vehicle: a vehicles outline fits into a rectangle, this rectangle is sized to allow vehicles to use local roads, and rectangles from two different vehicles are disjoint. Our shape-first approach can efficiently process high-resolution overhead imaging over wide areas to provide tips and cues for human analysts, or for subsequent automatic processing using machine learning or other analysis based on color, tone, pattern, texture, size, and/or location (shape first). In fact, computationally-intensive complex structural, syntactic, and statistical analysis may be possible when a shape-first work flow sends a list of specific tips and cues down a processing pipeline rather than sending the whole of wide area imaging information. This data flow may fit well when bandwidth is limited between computers delivering ad hoc image exploitation and an imaging sensor. As expected, our early computational experiments find that the shape-first processing stage appears to reliably detect rectangular shapes from vehicles. More intriguing is that our computational experiments with six-inch GSD OIRDS benchmark images show that the shape-first stage can be efficient, and that candidate vehicle locations corresponding to features that do not include vehicles are unlikely to trigger tips and cues. We found that stopping with just the shape-first list of candidate vehicle locations, and then solving a weighted, maximal independent vertex set problem to resolve conflicts among candidate vehicle locations, often correctly traces the vehicles in an OIRDS scene.

HYDROCODE DEVELOPMENT ON THE NCUBE AND THE CONNECTION MACHINE HYPERCUBES

Hydrocode simulations constitute an important tool at Sandia National Laboratories and elsewhere for analyzing complex two- and three-dimensional systems. However, current vector supercomputers do not provide a growth path to enable fast, routine, and cost-effective simulations of large problems. Massively-parallel computers will provide a solution. Sandia has already developed simplified versions of the production hydrocode CTH for the Connection Machine and the nCUBE massively-parallel supercomputers. The parallel versions solve problems in two-dimensional, multi-fluid, shock-wave physics. Code development strategy, coding methodology, visualization techniques and performance results for this work are described.

Practical, computer-aided registration of multiple, three-dimensional, magnetic-resonance observations of the human brain

We define a methodology for aligning multiple, three-dimensional, magnetic-resonance observations of the human brain over six degrees of freedom. The observations may be taken with disparate resolutions, pulse sequences, and orientations. The alignment method is a practical combination of off-line and interactive computation. An off-line computation first automatically performs a robust surface extraction from each observation. Second, an operator executes interactively on a graphics workstation to produce the alignment. For our experiments, we were able to complete both alignment tasks interactively, due to the quick execution of our implementation of the off-line computation on a highly-parallel supercomputer. To assess accuracy of an alignment, we also propose a consistency measure.

Out-of-core backpropagation

Backpropagation learning can execute at supercomputer speed from training data sets of unprecedented size when supercomputer main memory is backed with newly available parallel arrays of commodity disk drives. An efficient implementation of backpropagation learning was modified and extended to iterate through training data sets stored on a parallel-disk backing store. The algorithm is standard, including generating and adding noise to training inputs and the usual momentum term. With data sets up to the 10-GB capacity of the backing store, this backpropagation executes on the 16384-processor Connection Machine (CM) and parallel disks at 9.3 million connections per second (MCPS). A rate of 31 MCPS on a 65536-processor CM is predicted. This backpropagation reliably and automatically trained a feedforward, two-hidden-layer artificial neural network classifier with 33824 weights using 67584 input-output training pairs. This practical, forward-and-backward training computation through one pass of the data executed in 4.1 min. Moving the training from data parallel disks to main memory took 6.5% of the execution time. On a 65536-processor CM, a time of 74 s, with 22% spent on this data movement, is projected