Philippe Borianne

A Forest Growth Simulator Based on Functional-Structural Modelling of Individual Trees

The objective of this paper is to study forest growth simulation based on functional-structural modelling and its potentials for forestry applications. The GreenLab model is used for this purpose owing to its computational performances, its calibration capacity on real plants and its extension to the stand level, by taking into account the competition between neighbouring plants and the interactions with the environment. We first propose a software design: - to manage the composition of forest scenes, - to simulate their growth based on parallel computing of individual trees with the GreenLab model, - to get the realistic and real-time 3D rendering of the simulation results. We then detail a test case to illustrate the potentialities of this new tool. Mono-specific stands of poplars and pines are simulated. We analyze the computation performances and illustrate the simulation results with 3D outputs. A very classical application in forest management, stand thinning, is also tested. Our tool provides new insights thanks to the detailed architectures of trees resulting from the functional-structural model.

Automated delineation of tree-rings in X-Ray Computed Tomography images of wood

In this paper, the authors present an original method for the automated recognition of tree-ring limits in X-Ray Computed Tomography images of wood. The method is based on two steps: tree-ring tagging, and tree-ring delineation. The first step is performed by analyzing the radial intensity profiles after locating automatically the pith, and the second step uses active contours which iteratively detect the different tree-rings, from the bark to the pith. In particular, the geometrical constraints included in the active contour algorithm allow getting coherent limits, without break or discontinuity. The method is robust enough to override the main artifact met in dendrology, as for example, nodes or splits and to deal with the very high variability of different tree species.

Automatic characterization of the cell organization in light microscopic images of wood: Application to the identification of the cell file

Automated analysis of wood anatomical sections is of great interest in understanding the growth and development of plants. In this paper, we propose a novel method to characterize the cell organization in light microscopic wood section images. It aims to identify automatically the cell file in a context of mass treatment. The originality of the proposed method is our cell classification process. Unlike many supervised methods, our method is self conditioned, based on a decision tree which thresholds are automatically evaluated according to specific biological characteristics of each image. In order to evaluate the performances of the proposed system and allow the certification of the cell line detection, we introduced indices of quality characterizing the accuracy of results and parameters of these results. Those are related to topological and geometrical characters of the cell file at both global and local scales. Moreover, we propose an index of certainty for selective results exploitation in further statistical studies. The proposed method was is implemented as a plugin for ImageJ. Tests hold on various wood section well contrasted images show good results in terms of cell file detection and process speed.

Automatic identification of cell files in light microscopic images of conifer wood

In this paper, we present an automatic method to recognize cell files in light microscopic images of conifer wood. This original method is composed of three steps: the segmentation step which extracts some anatomical structures from the image (based on the watershed algorithm), the classification step which identifies the interesting cells in these structures (by using CART method), and the cell files recognition step. Some preliminary results obtained on several species of conifers are presented and analyzed.

An automated method for tree-ring delineation based on active contours guided by DT-CWT complex coefficients in photographic images

Our method can be used to delineate successive concentric tree-rings in Abies alba images.We coupled the Q-Shift DT-CWT and the points-based active contour model.Notions such as attraction power and orientation simulate an external force field.Our method performed as well as experts delineations on 200 tree-rings. This paper describes an efficient method for delineating tree-rings and inter tree-rings in wood slice images. The method is based on an active contour approach and a multi-scale gradient map resulting from the Dual Tree Complex Wavelet Transform (DT-CWT). The method is automated and does not require any pith localization. It is also quite robust to some defect structures such as branch prints, cracks, knots or mold. We applied the method to process entire Abies alba wood slices (aged from 10 to 50years) from bark to pith, which amounted to about 200 tree-rings. Our automatic delineation method performed accurately compared to the manual expert measurements with a mean F-score of 0.91 for the quality of delineation.

Simple-Graphs Fusion in Image Mosaic: Application to Automated Cell Files Identification in Wood Slices

Results aggregation by disjoint graph merging is potentially a good alternative to image stitching. During the processing of image mosaics, it allows to be free of radiometric and geometric corrections inherent in image fusion. We have studied and developed a generic merging method of disjoint graphs for tracking cell alignments in image mosaics of wood.

Automated valuation of leaves area for large-scale analysis needing data coupling or petioles deletion

In this paper, we present a new fully-automated approach to analyze the geometry of homogeneous leaves from numerical images produced by a desktop scanner. Our method is focused on two particular aspects: data coupling, specifically weights and areas matching, and numerical deletion of petioles, both in a large-scale analysis context. The process limits user interaction, the leaves segmentation is self conditioned and realized by a k-means classification giving the threshold value for each image. The results are arranged according to the leaves disposition on the scanner pane: the leaves labels are defined by recognitions of rows, obtained by clustering. The result layout eases the data coupling, i.e. the matching of individual leaves measures produced by different devices. We do also propose a new automated petiole removal procedure. The deletion method, applied to each single leaf, uses a morphological analysis from the leaf medial line. The process is adaptive, without expert calibration. The method is implemented in a ready to use application. Tests hold on several data sets show satisfactory results on a wide range of leaf shapes.

Fast Semi-supervised Segmentation of in Situ Tree Color Images

In this paper we present an original semi-supervised method for the segmentation of in situ tree color images which combines color quantization, adaptive fragmentation of learning areas defined by the human operator and labeling propagation. A mathematical morphology post-processing is introduced to emphasize the narrow and thin structures which characterize branches. Applied in the L*a*b* color system, this method is well adapted to easily adjust the learning set so that the resultant labeling corresponds to the accuracy achieved by the human operator. The method has been embarked and evaluated on a tablet to help tree professionals in their expertise or diagnosis. The images, acquired and processed with a mobile device, present more or less complex background both in terms of content and lightness, more or less dense foliage and more or less thick branches. Results are good on images with soft lightness without direct sunlight.

GT-RootS: An integrated software for automated root system measurement from high-throughput phenotyping platform images

GT-RootS (Global Traits of Root System) is an automated Java-based open-source solution we are developing for processing root system images provided by the Rhizoscope, a CIRAD phenotyping platform dedicated to dense cereal plants. Two types of use are proposed. The fully-automated mode applies a predefined standard processing pipeline to a preselected set of images while the semi-automated mode allows the user to interactively check and correct intermediate processing results to a specific image. In both cases, GT-RootS combines a local adaptive thresholding algorithm and a similarity indicator to automatically separate the root system from a complex background without user intervention. A covering house-shaped polygon is then defined in the axis system of the root ellipse from vertical weighted density profiles. This canonical shape is composed of both upper trapezoid and lower rectangular compartments from which upper and lower heights, global width and local offset, root system cone angulation and spatial densities can be easily evaluated and displayed. GT-RootS measurements were compared both to expert evaluations and to two other estimation methods on a set of 64 images of a dense Japonica rice root system of 30-days-old plants. We demonstrate also that GT-RootS satisfies the requirements of high-throughput analyses: short processing time (around 30 images per hour on a low-end computer), measurement accuracy and repeatability, and user bias eradication.

Automated efficient computation of crown transparency from tree silhouette images

The transparency of trees is the most important indicator for a forest health assessment. This paper presents an efficient method for calculating the crown transparency coefficient from tree binary images. This coefficient is based on the automated quantification of the deep indentation, macro-hole and micro-hole densities. Circular structuring elements are introduced, among other things, to automatically find the significant biological size. The symmetric tree convex hull and the tree smoothed contour are defined to delineate the reference areas necessary to evaluate the above-mentioned densities. Statistical thresholds are proposed to eliminate human operator subjectivity, especially in the automated identification of anatomical elements such as soft and deep crown-indentations or macro and micro crown-holes. A point-wise transparency map is produced to better appreciate the origin of the visible skylight areas in the crown. The crown micro-hole density is calculated from the 0.1-to-0.5 transparency points, the crown macro-hole density from the 0.5-to-1 transparency points. We finally opt for weighting of the above three densities with regard to the importance of the symptoms they describe for a more relevant crown transparency coefficient. A comparative study on several trees from full-size and half-size binary images showed that our method is similar overall to the DSO and less sensitive to scale reduction.