Key message High-resolution imaging is possible if high-frequency sensors are used together with a signalprocessing and inversion algorithm that is well suited to a low signal-to-noise ratio and the effect of wood anisotropy. Wood is a biological growth medium, and given that standing trees adapt themselves in their growth to environmental conditions, their material properties vary with age. These changes result in variations that are far more complex than anisotropy. Wood quality and intraspecific variability can thus be studied to gain an understanding of the development mechanisms of trees, and this can be useful for clonal selection and the management of tree communities. A number of techniques are available to determine wood properties in standing trees, but the signal-processing approaches currently used are not always robust and do not always provide the image resolution needed in the particular cases of acoustic or ultrasonic tomography. This review paper thus aims to present important aspects that should be taken into account when using tomography techniques and addresses a number of open problems. A brief review of current non-destructive wood imaging techniques is initially presented followed by a comparison of the protocols, methods and models used in acoustic and ultrasonic tomography. The devices cited were studied in terms of measurement systems and signal processing. The analysis aimed to highlight and analyze the advantages and disadvantages of each device and describe challenges and trends. The effect of various parameters is discussed: frequency, signal-to-noise ratio, number of sensors and inversion algorithm. General conclusions are then drawn in relation to future signal-processing work in the acoustic and ultrasonic tomography of standing trees. (Résumé d'auteur

No abstract

Context The quality of acoustic tomographic images in standing trees is mainly function of the accuracy of the acous- tic velocity computation. Improving the acoustic velocity determination is, furthermore, of great interest because acous- tic tools are widely used in nondestructive testing of wood. • Aims Four different signal processing algorithms were used (1) to study the effect of the signal dynamic on the velocity determination, (2) to determine the validity range of each computation method, and (3) to compare the behavior be- tween a homogeneous material and wood. • Methods The experiments were performed using the con- ventional experimental protocol for the ultrasonic characteri- zation of materials in a tank (normal incidence transmission at 500 kHz). A polyurethane resin (homogeneous material) and two wood species ( Bagassa guianensis Aubl., Milicia excelsa (Welw.) C.C. Berg) were used for the experiments. • Results Computed velocity increased as the noise level increased. The Hinkley method appeared to be the most exact when the noise level exceeded 10 dB. The Fisher method was that most suitable for very noisy signals. No difference was found between the resin and the wood samples. • Conclusion A combination of the Fisher and Hinkley methods in the same algorithm would yield the most accurate acoustic velocity determinations in the tomography of stand- ing trees. Key message Wood acoustic velocity determination is affect- ed by the wavelength and the detection algorithm used. The Fisher algorithm is optimal with high signal attenuation; oth- erwise, the Hinkley algorithm should be used. (Résumé d'auteur

International audienceThe quality of the Ultrasonic Computed Tomography (UCT) imaging device, in term of spatial resolution, sensitivity or dynamic, is a function of the wavelength, the number of transducers, the ultrasonic field generated and the inversion algorithm. This latter factor was studied using four inversion techniques for the tomography of trees. Two new methods, Partial Least Squares method - PLS, and Layer Stripping - LS, are compared to two classical methods, Filtered Backprojection - FBP, and Simultaneous Iterative Reconstruction Technique - SIRT. An original numerical phantom of tree was used and the effects of the variation of the number of transducers and the noise level were analyzed. The PLS was found to converge slowly but reached a high performance for the reconstruction of the projection values (slowness). PLS method was also characterized by a fast computation time. However, the PLS method gave a poor image quality (high sensitivity to noise level and presence of outlier values). The main advantage of the LS method was its high robustness to the noise level. The computing time was however the weakest point of LS method. A combination of the two classic FBP-SIRT methods seemed the most suitable except for the computation time. However, an improvement of the convergence conditions of PLS method (very fast computing time; including also curved rays), with the use of SIRT method, would constitute a promising solution

In the assessment of standing trees, an acoustic tomographic device is a valuable tool as it permits to acquire data from the inner part of the trees without causing them to fall down unnecessarily. The interpretation of the images produced by these devices is part of the diagnosis process for urban trees management. This paper presents a segmentation methodology to identify defective regions in cross-section tomographic images obtained with an Arbotom® device. Two trunk samples obtained from a Blackwood Acacia tree (Acacia melanoxylon) were tested, simulating defects by drilling holes with known geometry, size and position and using different numbers of sensors. Tomograms from the trunk cross sections were processed to align the propagation velocity data with the corresponding region, either healthy or defective. The segmentation methodology proposed aims to find a velocity threshold value to separate the defective region adjusting a logistic regression model to obtain the value that maximizes a performance criterion, using in this case the geometric mean. Two criteria were used to validate this methodology: the geometric mean and the surface ratio detected. Although an optimal threshold value was found for each experiment, this value was strongly influenced by the defect characteristics and the number of sensors. The correctly segmented area ranging from 54 to 93% demonstrates that the threshold method is not always the most proper way to process this type of images, and thereby further research is required in image processing and analysis. (Résumé d'auteur