A digital blurred plane is a finite set of points with controlled thickness. This notion can be used to characterize a real world object surface, considered as planar at a large scale, but actually quite irregular at finer scale. In this paper, we introduce a new framework to detect a blurred plane in a point set, based on the analysis of two orthogonal height profiles using digital geometry tools. It is applied to the interactive extraction of relict charcoal platforms from an airborne LiDAR point cloud. This process was tested on two sets of geolocated platforms obtained by on-site prospecting or visual survey of the LiDAR digital terrain model. Experimental validations show that a large amount of them is successfully extracted and that their estimated extent closely matches the ground truth.

Accurate location of access roads is important for forest management, in particular in mountain areas. In this paper, we are interested in their detection from LiDAR data using deep learning approaches. For this, we use images computed from an interpolated surface, called digital terrain model (DTM), of the 3D point cloud. In order to train and validate the neural network models, two ground truth datasets associated to DTM images are considered: (1) manual digitization of the road centerlines and (2) automatic extraction followed by supervised completion using two softwares based on discrete geometry tools. The trained network models are then evaluated over a test dataset using standard measures such as precision, recall, F-measure and prediction time.

The SolHoM(a) project aims at determining the degree of landscape anthropisation in the Fossard massif (Remiremont, Vosges, France) by estimating the impact of past human occupations on current soils. Nowadays, a forest covers this relatively inhospitable mountain (hard climate, steep slopes, acid soils ...), but some remains suggest human presence for at least 1400 years. The Saint-Mont is a variscan orogenic granitic summit (height 672m) with valuable preserved archaeological relics, located south of the Fossard. At the top was founded the monasterium Habendum (from the 7th century to the French Revolution), which potentially comes after a late-antic castrum. Furthermore, monumental enclosure dry-stones walls, not yet clearly linked with the monastery, crosses the massif. Our methodology is first experienced on the Saint-Mont, well documented by recent archaeological research. It features: - sampling soil transects presenting a clear gradient from highly to poorly human impacted areas. According to those transects, soils are sampled and characterised (physico-chemistry, soil micromorphology, analysis of organic matter, dating); this should highlight traces of past human activity; - a development of algorithms to detect geometric structures in the raw LiDAR data provided by the PCR AGER(b) (Ch. Kraemer, dir). This research might help archaeologists to determine areas influenced or not by men. Once tested on the Saint-Mont, this approach will be applied to another archaeologically less investigated area of the Fossard. First field works highlight two soil sequences: - the first one, near a probable merovingian wall close to the monastery, reveals a very black 80cm deep unit, - the second one, far-off the monastery and beside the enclosure dry-stone wall, is less thick but have a thin black unit buried under 20 cm of pedogenised sediment. Physico-chemical and soil micromorphological analyses are in progress. The coming study of profiles/sequences more distant from archaeological sites will contribute to estimate the degree of soils anthropisation in the Saint-Mont. (a) Interactions Sol-Hommes-Milieux (b) ArcheoGEographie du premier Remiremont et de ses abords

In this paper, a framework is proposed to extract forest roads from LiDAR (Light Detection and Ranging) data in mountainous areas. For that purpose, an efficient and simple solution based on discrete geometry and mathematical morphology tools is proposed. The framework is composed of two steps: (i) detecting road candidates in DTM (Digital Terrain Model) views using a mathematical morphology filter and a fast blurred segment detector in order to select a set of road seeds; (ii) extracting road sections from the obtained seeds using only the raw LiDAR points to cope with DTM approximations. For the second step, a previous tool for fast extraction of linear structures directly from ground points was adapted to automatically process each seed. It first performs a recognition of the road structure under the seed. In case of success, the structure is tracked and extended as far as possible on each side of the segment before post-processing validation and cleaning. Experiments on real data over a wide mountain area (about 78 km2) have been conducted to validate the proposed method.

Abstract. In this paper, a general framework is proposed for live extraction of curvilinear structures such as roads or ridges from airborne LiDAR raw data, in the scope of present and past man-environment interaction studies. Unlike most approaches in literature, classified ground points are directly processed here, rather than derived products such as digital terrain models (DTM). This allows to detect possible lacks of ground points due to LiDAR signal occlusions caused by dense coniferous canopies. An efficient and simple solution based on discrete geometry tools is described for supervised context in which the user just indicates where the extraction should take place. Fast response times are required to ensure a good man-system interaction.The framework performance is first evaluated on the example of the extraction of forest roads in a mountainous area, as these objects are well marked in the DTM and hence provide some kind of ground truth. Good execution time and accuracy level are reported. Then this framework is applied to the detection of prominent curvilinear structures, which are much more diffuse objects, but of greater interest than roads in the scope of the present project. Achieved results show high potential of the proposed approach to help archaeologists and geomorphologists in finding areas of interest for future prospection using LiDAR data.

This document presents complementary experiments on the published algorithm of Thick Line Segment Detection with Fast Directional Tracking. The main paper is actually published at ICIAP 2019 [2]. First tests compare the performance of the detector with and without adaptive directional scans (ADS) and assigned thickness control (ATC). On the detector without ADS, the ne tracking step must be performed twice to get less risk of growing blurred segment escape from the scan strip.

This paper introduces a fully discrete framework for a new straight line detector in gray-level images, where line segments are enriched with a thickness parameter intended to provide a quality criterion on the extracted feature. This study is based on a previous work on interactive line detection in gray-level images. At first, a better estimation of the segment thickness and orientation is achieved through two main improvements: adaptive directional scans and control of assigned thickness. Then, these advances are exploited for a complete unsupervised detection of all the line segments in an image. The new thick line detector is left available in an online demonstration.

The segmentation or the geometric analysis of specular objects is known as a difficult problem in the computer vision domain. It is also true for the problem of line detection where the specular reflection implies numerous false positive line detection or missing lines located on the dark parts of the object. This limitation reduces its potential use for concrete industrial applications where metallic objects are frequent. In order to overcome this limitation, a new strategy to detect thick segment is proposed. It is not based on the image gradient as usually, but rather exploits the image intensity profile defined inside a parallel strip primitive. Associated to a digital straight segment recognition algorithmwhich is robust to noise, this strategy was implemented to track metallic tubular objects in gray-level images. The efficiency of the proposed method is demonstrated through extensive tests using an actual industrial application. An alternate release intended to overcome the possible impact of the digitization process on the achieved performance is also introduced. Both strategies are discussed at the end of the article.

The segmentation or the geometric analysis of specular object is known as a difficult problem in the computer vision domain. It is also true for the problem of line detection where the specular reflection implies numerous false positive line detection or missing lines located on the dark parts of the object. This limitation reduces its potential use for concrete industrial applications where metallic objects are frequent. In this work, we propose to overcome this limitation by proposing a new strategy which is not based on the image gradient as usually, but exploits the image intensity profile defined inside a parallel strip primitive. Associated to a digital straight segment recognition algorithm robust to noise, we demonstrate the efficiency of our proposed method with a real industrial application.

The recognition of discrete straight segments is a significant topic in the field of discrete geometry and for many applications dealing with geometric feature extraction. It can be performed from noisy binary data using the concept of blurred segments [3,2]. However, to our best knowledge, these algorithms have never been defined to directly extract straight segments in gray level images. This article proposes a solution to extend the recognition by using gray level image information. Although initially intended to be implemented within a semi-automatic line selection tool used in an interactive 3D modeling application, it also meets more general parameter extraction requirements.

This paper describes a model of three-dimensional cellular automata allowing to simulate different phenomena in the fields of computer graphics or image processing. Our method allows to combine them together in order to produce complex effects such as automatic texturing, surface imperfections, or biological retina multi-layer cellular behaviours. Our cellular automaton model is defined as a network of connected cells arranged in a natural and dynamic way, which affords multi-behavior capabilities. Based on cheap and widespread computing systems, real-time performance can be reached for simulations involving up to a hundred thousand cells. The efficiency of such an approach is illustrated by a set of CA related to computer graphics –e.g. erosion, sedimentation, or vegetal growing processes– and image analysis –e.g. retina simulation.

Architectural sites suffers from weathering effects such as liquid sedimentation, lampblack or pollution deposit, moss growing or longterm erosion. Displaying their impact on present buildings could help deciders to plan relevant actions for preserving the architectural heritage. In this paper we present a user-friendly solution for on-line acquiring a semi-physical model from one or few images of a building, immediately followed by real-time simulations. The goal is certainly not to give an accurate prediction of what will happen, but rather to show a plausible state of the site under various possible degradations arising from natural effects. The solution relies on the coupling of two 3-D modelling programs: one dedicated to interactive reconstructions and a second to discreet manipulation for natural phenomena simulations based on a combination of different surface cellular automata. Several experiments were held to estimate the effectiveness of this approach and to identify future works in order to cope with present limitations. A first series applies a lampblack deposit simulation on a large-size 3-D model to demonstrate the solution feasibility and to provide indications about achievable performance. The second series focuses on the solution sensitiveness to both geometrical and semi-physical models granularity through the application of a stone erosion effect on a smaller but less regular-shaped building model. The third series demonstrates the discreet model ability to simulate multiple effects with the example of a vegetation growth simulation. Guidelines for future works are given in the conclusion of this paper.

This paper presents a generic and user-friendly procedure for interactive 3-D model registration on images. Such tasks are numerous in a CAD-based model acquisition process through interactive means. Moreover object alignment can also be used to quickly provide an initial match when registering a whole model on actual data. This task could be performed by setting a minimal set of 2D-3D point matches. Our goal is to set up a more user-friendly solution, which minimizes the amount of manual work. We have implemented a two-step procedure relying on the selection in the image of some structured object features. First a visual feature interpretation method is used. It provides a solution that fixes at least two rotation parameters. Remaining degrees-of-freedom are then easily handled through interactive modes. Several interpretation methods are proposed to cope with various contexts according to the available features. Achieved accuracy is a posteriori estimated to help the operator decide on the registration reliability.

Interactive methods are well suited to telerobotics purposes. Based on the super-imposition of the model on video images, they provide a friendly way to acquire or update the environment model from an on-board CCD camera. It is a flexible way to cope with a priori model inaccuracies and any possible mission incidents. Modelling time is a key requirement which can be fulfilled through the integration of semi-automatic assistances. This paper deals with the semi-automation of edge segment specification tasks. Edge segments combinations define 3-D lines or planes, which are used to orient the modelled objects. Their accurate specification is a tiring and time consuming task, which can not be fully automated. In the semi-automatic mode we set up, the operator quickly draws a line over the image. This line is automatically attracted towards the nearest extracted edge contour. The automatic attraction function is based on the Hough transform. This semiautomatic assistance has been integrated into the Pyramide interactive 3-D modelling system, which has been developed at CEA/STR. Evaluations on realistic sites showed its high flexibility and efficiency.

Environment 3D models provide efficient assistance to the supervision of telerobots. Modeling tools are necessary to acquire them in a compatible time with the mission requirements. We already developed an interactive modeling system which relies on images provided by an embedded camera. In order to reduce the modeling time, structural knowledge on the environment is integrated within specialized workshops, including dedicated primitives, constraints and modeling methods. We have demonstrated this approach with the example of a piping workshop, an application developed in collaboration with COGEMA and representative of a real environment in a nuclear fuel reprocessing facility

Virtual reality techniques are a promising way to enhance the supervision of a Computer Aided Teleoperation system. We already showed that interactive 3-D modeling is a flexible and sometimes the only possible way to provide the required environment model. In this article we indicate how a combined approach for model acquisition and exploitation can lead to an optimized implementation of both steps, and demonstrate it in the scope of a realistic maintenance mission: the inspection of a tap using a Maestro hydraulic manipulator.

Describes a telerobot control principle based on a close cooperation between the man and the machine. The main feature is the management of a 3D geometrical model of the environment through successive modelling-programming-execution sequences. The principle was applied to the control of a quadrumane climbing robot. During this experimental validation phase, the authors used the interactive 3D modelling system Pyramide to acquire the 3D environment model, to program actions, and to monitor the robot motions. The authors give details of the experiments realized.<>