Algorithm for automatic detection and analysis of cracks in timber beams from LiDAR data

Highlights

• Natural wood generally cracks when it dries.

• Mapping beam cracks is a necessary task to evaluate their potential danger.

• The new algorithm proposed is used to identify, analyze and label each crack.

• Each crack is analyzed based on the points that define its edge.

• A data file with all geometric characteristics of each identified crack is obtained.

Abstract

Cracks may indicate structural problems in constructions made of timber beams, but their identification in notably roof constructions is difficult. Therefore this paper presents an algorithm for the automatic detection of cracks in timber beams sampled by LiDAR data. This algorithm enables the identification, analysis and monitoring of cracks and their geometrical characteristics. The algorithm is validated by tests on laboratory specimens and on a timber roof structure. The results prove that the proposed algorithm is able to automatically identify and analyze cracks with a width of above 3 mm from point clouds with an average resolution below 1 mm.

Introduction

Along history, wood has been a material commonly used in the construction of all kind of buildings and structures, such as houses, bridges, towers, windmills, etc. In fact, nowadays there are innumerable historical timber structures around the world. To preserve this heritage, its conservation and maintenance is needed. For this reason, structural health control is a fundamental issue. Also for the rehabilitation of old buildings for alternative use, analysis of structural health is mandatory. During the last century the use of wood as material for structures has declined in favor of other materials such as steel or concrete. Nevertheless, in recent decades the use of both solid and engineered laminated timber in structures has increased again.

Natural wood generally cracks when it dries. This is caused by the anisotropic behavior of wood in radial versus tangential direction along the cross-section of the beam (Fig. 1a). Cracks are present in both old and new structures. For new timber structures mainly glue laminated timber is used. Cracks may develop within the timber lamella or in the glue line and this is generally caused by stresses perpendicular to the grain as a result of temperature and wood moisture fluctuations in the beam (Fig. 1b). The loading and the duration or variation of loading also influence crack growth, especially in areas with complex stress states.

Many cracks are harmless, but other cracks negatively affect the strength of a structure. Depending on the position, size and shape of the crack, it may have consequences for the safety of the structure, which in addition depends on the beam loading (Fig. 1e). For example, large cracks in the neutral axis of a beam decrease the load bearing capacity of the timber element in shear and can also result in a new stress distribution which may result in crack propagation. Cracks following the grain (fibre) that deviate from the neutral axis need to be evaluated depending on their size and location (Fig. 1d). Mapping beam cracks is a necessary task to evaluate their potential danger. The potential impact of a crack could be determined after crack detection using finite element analysis. Mapping all cracks in for example a wooden roof structure using contact measurements is time consuming and often challenging, because of the height of the structure above the floor. The importance of the depth of a crack for solid wood strength properties has been investigated by Frech [1]. High growth rates, the presence of pits, knots and juvenile wood in the beam increases the hazard of developing cracks [2]. However for service life prediction and safety considerations, controlling and evaluation of arising and growing cracks are essential.

For initial inspection and control of building and structures, non-destructive methods are most suitable because the material is not, or very limited, damaged by the measurements. Examples of such methods are: acoustic methods, resistance drilling techniques and radiography [3], [4], [5]. Also some inspection work has been performed using Ground Penetrating Radar (GPR) [6]. These methods allow a detailed analysis of the health of a timber structure but, they have the disadvantage that it is necessary to access the object and make contact with it.

Photogrammetry is a technique which enables health monitoring without direct contact of the scene. Several researchers have applied photogrammetry to assess cracks. Barazzetti and Scaioni [7] present an image-based method for crack analysis; Sohn et al. [8] propose a crack-monitoring system to quantify the change of cracks in concrete from multitemporal images. Patricio and Maravall [9] present a novel generalization of the gray-scale histogram and its application to the automated visual measurement and inspection of cracks in wooden pallets. Although photogrammetry is in general a suitable technique to assess cracks, photogrammetry also depends on the light, color and texture of the specimen surface to analyze. In some indoor situations bad lighting conditions may therefore hamper the use of photogrammetry for crack analysis.

Laser scanning is a promising alternative technique for structural health analysis [10], [11], [12]. Laser scanning provides several advantages with respect to other methods such as: enabling the analysis of inaccessible structural elements, obtaining the 3D surface without direct contact to the specimen, efficient data collection in the form of ready to use 3D point clouds, no dependency on lighting conditions, color and texture of the specimen surface under analysis. Some applications of laser scanning for structural analysis are: dimensional compliance control [13], construction progress control [14], analysis and documentation of historic structures [15], [16], 3D modeling for structural engineering purposes [17], [18], [19] or analysis of structural health [11], [12].

In the field of timber structures, laser scanning has been successfully used for 3D modeling of historic structures. Alessandri and Mallardo [20] deals with some preliminary structural analysis of the wooden roof construction of the Church of the Nativity in Bethlehem; Balletti et al. [21] study the wooden roof structure of the dome of SS. Giovanni e Paolo in Venice; Oreni et al. [22] illustrates the switch from a 3D content model to a Historic Building Information Model (HBIM) in order to support conservation and management of built heritage. However, to the best of our knowledge only Van Goethem et al. [23] has reconstructed and modeled timber boards with knots from LiDAR data. Knots and their location in the boards were analyzed and structural wood quality could be estimated.

Some work has been performed on the analysis of cracks sampled by LIDAR data in non-wooden structures. Laefer et al. [24] presents the fundamental mathematics to determine the minimum crack width detectable with a terrestrial laser scanner in unit-based masonry; Laefer et al. [25] compares the relative performance capabilities of crack detection by sidewalk-based manual inspection with digital photography. Sánchez-Aparicio et al. [27] presents a set of procedures based on laser scanning, photogrammetry and operational modal analysis to improve calibration of finite element models in historical buildings. This work includes cracks in masonry walls. In the same way, Huang et al. [26] present a pavement crack detection method combining 2D and 3D information based on Dempster‐Shafer Theory. Nevertheless, none of these works address the automatic detection and quantification of cracks in wood structures from LIDAR data.

The aim of this paper is to develop a first algorithm for the automatic detection of cracks in timber beams from LiDAR data. This algorithm allows analyzing the geometric characteristics of each identified crack, as well as generating a data file that labels each crack including all of its geometric characteristics. This algorithm will in addition be used for monitoring the possible growth of each crack in time. After this introduction, Section 2 outlines the proposed and developed algorithm. In Section 3 a series of tests with laboratory specimens and real timber structures are performed to check the algorithm and experimental results are presented. In Section 4 results and applicability of the proposed method are discussed. Finally, in Section 5, conclusions are given.