Clumping index (CI) is a measure of foliage aggregation relative to a random distribution of leaves in space. The CI can help with estimating fractions of sunlit and shaded leaves for a given value of leaf area index (LAI). Both the CI and LAI can be obtained from global Earth Observing (EO) systems such as the Moderate Resolution Imaging Spectrometer (MODIS). Here, the compatibility between CI and LAI products derived from EO data is examined independently using the theory of spectral invariants, also referred to as photon recollision probability theory (i.e. ` p-theory'), along with raw LAI-2000/2200 Plant Canopy Analyzer data from 75 sites distributed across a range of plant functional types (PFTs). The p-theory describes the probability (p-value) that a photon, having intercepted an element in the canopy, will recollide with another canopy element rather than escape the canopy. Our results indicate that the integration of empirically-based CI maps with the MODIS LAI product is feasible, providing a potential means to improve the accuracy of LAI EO data products. Given the strong results for the large range of PFTs explored here, we demonstrate the capacity to obtain p-values for any location solely from EO data. This is relevant for future applications of the photon recollision probability concept for global and local monitoring of vegetation using EO data.

We describe a parameterized memory system suitable as target for automatic high-level language to hardware compilers for reconfigurable computers. It fully supports the spatial computation paradigm by allowing the realization of each memory operator by a dedicated hardware memory port. Inter-port coherency is maintained only for those ports that actually require it, and efficient speculative execution is enabled by a dynamic scheme for arbitrating access to shared resources (such as main memory), relying on techniques inspired by the branch prediction of conventional software-programmable processors.

We propose an execution model that orchestrates the fine-grained interaction of a conventional general-purpose processor (GPP) and a high-speed reconfigurable hardware accelerator (HA), the latter having full master-mode access to memory. We then describe how the resulting requirements can actually be realized efficiently in a custom computer by hardware architecture and system software measures. One of these is a low-latency HA-to-GPP signaling scheme with latency up to 23× times shorter than conventional approaches. Another one is a high-bandwidth shared memory interface that does not interfere with time-critical operating system functions executing on the GPP, and still makes 89 percent of the physical memory bandwidth available to the HA. Finally, we show two schemes with different flexibility/performance trade-offs for running the HA in protected virtual memory scenarios. All of the techniques and their interactions are evaluated at the system level using the full-scale virtual memory variant of the Linux operating system on actual hardware.

Geometric algebra (GA) is a mathematical framework that allows the compact description of geometric relationships and algorithms in many fields of science and engineering. The execution of these algorithms, however, requires significant computational power that made the use of GA impractical for many real-world applications. We describe how a GA-based formulation of the inverse kinematics problem from computer animation and robotics can be accelerated using reconfigurable FPGA-based computing and using a graphics processing unit (GPU). The practical evaluation covers not only the sheer compute performance, but also the energy efficiency.

Adaptive computers combine conventional software programmable processors with reconfigurable compute units. We present techniques that allow the high-performance realization of demand-paged, virtually addressed main memory shared between both of these processing elements. Furthermore, we have achieved low-latency communication between software running on the CPU and the reconfigurable compute unit, allowing even fine-grained hardware/software partitioning. A system-level evaluation quantifies the advantages of our approach.

Insect-induced damages in forests are a major concern for timber production, landscape conservation and ecosystem research. Early detection methods based on remote sensing data can document the severity and spatial extent of ongoing attacks and might aid in designing mitigation measures or even prevention where necessary. In southeastern Norway, a large-scale insect defoliation of pine trees is ongoing. The larvae of the Pine sawfly Neodiprion sertifer create it with its mass attacks during their feeding on needles in June and July. In the winter before the attack, egg galleries are evident in the needles. This provides a test case for early detection methods and remote sensing techniques for monitoring forest health. In the context of an ongoing project on REMote sensing of FORest health (REM-FOR) in Norway, we approach this problem by mapping leaf area index (LAI) before and after the attack in a test area (size around 20 km2). LAI is used as a proxy for the crown density, and decreasing trends not related to phenology (non-periodic) indicate defoliation. Estimates for effective LAI for two different years, 2005 and 2007, were derived using airborne laser scanning (LIDAR) calibrated with ground-based point measurements with the LAI-2000 Plant Canopy Analyzer (LI-CORreg, USA). These estimates are based on an application of the Beer-Lambert law and a threshold-based separation of laser reflections from the ground and from the canopy. We also obtained airborne high-resolution hyperspectral images (HySpexTM, Norsk Elektrooptikk, Norway) for the same reference years to investigate the spectral response of the affected forest. The set of two cameras deliver 330 different spectral channels in the wavelength range 400 to 1800 nm. The analysis might be done using advanced multivariate methods or spectral unmixing using spectral libraries. Here, it was performed using a physically-based model emphasizing geometrical and optical properties of canopies, the Forest Reflectance Model FRT. FRT was designed for the application to (managed) Northern European Forests and is based on conventional forest inventory data, species-dependent parametrized crown shapes, canopy LAI, needle clumping index, and needle optical properties. Here, however, we run the model in an inverse mode, by iteratively minimizing the discrepancy between measured and simulated reflectances, and predicting the LAI, keeping well-constrained parameters of the model (e.g. tree density, tree height, optical properties of needles) fixed while calibrating others (e.g. needle weight per tree, average shoot length, shoot self-shading). A set of 14 sample trees felled in 2005 further constrain the range of the calibrated parameters, excluding local model inversion minima with unrealistic parameter estimates. The LAI values predicted by the model are then compared to those obtained with airborne laser-scanning with a spatial resolution of 20 m times 20 m for the pine- dominated part of the scenes. In doing so, the spectra from each pixel (size 25 cm times 25 cm) were aggregated by calculating channel-wise median values. In effect, the modelling setup results in determining geometrical and optical properties of forest plots, trees and needles from hyperspectral images. The project data are complemented by images from a range of satellite-based sensors, including MODIS, SPOT, and Hyperion to cover larger regions and as a basis for operationalizing the approach for future insect attacks.

We introduce a new execution model for orchestrating the interaction between the conventional processor and the reconfigurable compute unit in adaptive computer systems. We then characterize the architectural and OS-level requirements of implementing the model, and demonstrate how they can be achieved on a real hardware platform running under a full scale multi-tasking virtual protected memory operating system. Experimental measurements show the efficiency of our solution, and also prove that reconfigurable computing can be competitive with processors even for non-streaming, pointer-chasing applications.

This work presents the realization of a general system of dense stereovision by simulated annealing. Dense stereovision is based on point matching. We show the performance of our approach compared to correlation. We use an optimization method in order to take into consideration the global aspect of the problem, as opposed to the correlation that acts locally on windows, and be able to incorporate this module with other early vision modules. The stereovision problem is an ill-posed problem where the global minimum is hidden by local minima and where the notion of gradient does not exist. For this reason, the simulated annealing algorithm seems the most suitable to solve the stereovision problem. The constraints of the stereovision problem are expressed by an energy function and elementary transformations.