This work addresses the reconstruction of strain gradient fields at the wood growth ring scale from full-field deformation measurements provided by digital image correlation. Moreover, the spatial distribution of the earlywood and latewood radial modulus of elasticity is assessed. Meso-scale tensile tests are carried out on Pinus pinaster Ait. wooden specimens oriented in the radial-tangential plane under quasi-static loading conditions. A parametric analysis of the twodimensional digital image correlation extrinsic and intrinsic setting parameters is performed, in a balance between spatial resolution and resolution. It is shown that the parametric module is an effective way to quantitatively support the choice of digital image correlation parameters in the presence of the high deformation gradient fields generated by the structure-property relationships at the scale of observation. Under the assumption of a uniaxial tensile stress state, the spatial distribution of the radial elastic modulus across the growth rings is obtained. It is observed that the ratio of the radial modulus of elasticity between latewood and earlywood tissues can vary significantly as a function of the digital image correlation parameters. It is pointed out, however, that a convergence value can be systematically established. Effectively, earlywood and latewood stress-strain curves are obtained and elastic properties are determined assuming the converged digital image correlation setting parameters.

This paper reports a novel application of hyperspectral imaging (a spectroscopic technique) for measuring wood density profi les at the growth ring scale. The measurements were performed with a spatial resolution of 79 μm. In the present case, hyperspectral imaging was used to measure wood sample refl ectance for light in the wavelength range between 380 and 1028 nm, with a resolution of approximately 0.6 nm. The work was performed with 34 samples collected from 34 trees of Pinus pinea . A total of 34,093 density points were used to create and validate a partial least-squares (PLS) regression that converts spectroscopic refl ectance data into density values. The coeffi cient of determination value between the present method and X-ray microdensitometry is 0.810 with a root mean squared error of 6.54 × 10 -2 g.cm -3 .

No abstract

Abstract-In this paper, we present a new method for neural spike sorting based on Continuous Time (CT) signal processing. A set of CT based features are proposed and extracted from CT sampled pulses, and a complete event-driven spike sorting algorithm that performs classification based on these features is developed. Compared to conventional methods for spike sorting, the hardware implementation of the proposed method does not require any synchronisation clock for logic circuits, and thus its power consumption depend solely on the spike activity. This has been implemented using a variable quantisation step CT analogue to digital converter (ADC) with custom digital logic that is driven by level crossing events. Simulation results using synthetic neural data shows a comparable accuracy compared to template matching (TM) and Principle Components Analysis (PCA) based discrete sampled classification.