Identification of materials from construction site imagesusing content based image retrieval techniques

摘要

The capability to automatically identify shapes, objects and materials from the imagerncontent through direct and indirect methodologies has enabled the development ofrnseveral civil engineering related applications that assist in the design, constructionrnand maintenance of construction projects. Examples include surface cracks detection,rnassessment of fire-damaged mortar, fatigue evaluation of asphalt mixes, aggregaternshape measurements, velocimentry, vehicles detection, pore size distribution inrngeotextiles, damage detection and others. This capability is a product of therntechnological breakthroughs in the area of Image and Video Processing that hasrnallowed for the development of a large number of digital imaging applications in allrnindustries ranging from the well established medical diagnostic tools (magneticrnresonance imaging, spectroscopy and nuclear medical imaging) to image searchingrnmechanisms (image matching, content based image retrieval).rnContent based image retrieval techniques can also assist in the automated recognitionrnof materials in construction site images and thus enable the development of reliablernmethods for image classification and retrieval. The amount of original imagingrninformation produced yearly in the construction industry during the last decade hasrnexperienced a tremendous growth. Digital cameras and image databases are graduallyrnreplacing traditional photography while owners demand complete site photographrnlogs and engineers store thousands of images for each project to use in a number ofrnconstruction management tasks. However, construction companies tend to storernimages without following any standardized indexing protocols, thus making thernmanual searching and retrieval a tedious and time-consuming effort. Alternatively,rnmaterial and object identification techniques can be used for the development ofrnautomated, content based, construction site image retrieval methodology. Thesernmethods can utilize automatic material or object based indexing to remove the userrnfrom the time-consuming and tedious manual classification process.rnIn this paper, a novel material identification methodology is presented. This methodrnutilizes content based image retrieval concepts to match known material samples withrnmaterial clusters within the image content. The results demonstrate the suitability ofrnthis methodology for construction site image retrieval purposes and reveal therncapability of existing image processing technologies to accurately identify a wealth ofrnmaterials from construction site images.