Changmin Kim

Fully Automated As-Built 3D Pipeline Extraction Method from Laser-Scanned Data Based on Curvature Computation

There has been a growing demand for the three-dimensional (3D) reconstruction of as-built pipelines. The as-built 3D pipeline reconstruction process consists of the measurement of an industrial plant, identification of pipelines, and generation of 3D models of the pipelines. Although measurement is now efficiently performed using laser-scanning technology, and in spite of significant progress in 3D pipeline model generation, the identification of pipelines from large and complex sets of laser-scanned data continues to pose a challenge. The aim of this study is to propose a method to automatically extract 3D points corresponding to as-built pipelines that occupy large areas of industrial plants from laser-scanned data. The proposed extraction method consists of the following steps: preprocessing, segmentation of the 3D point cloud, feature extraction based on curvature computation, and pipeline classification. An experiment was performed at an operating industrial plant to validate the proposed method. The experimental result revealed that the proposed method can indeed contribute to the automation of as-built 3D pipeline reconstruction.

Automatic 3D Reconstruction of As-Built Pipeline Based on Curvature Computations from Laser-Scanned Data

Demand has been growing for three-dimensional (3D) reconstruction of asbuilt pipelines that occupy large areas within operating plants. In practice, measurements are efficiently performed using laser-scanning technology; however reconstructing an as-built pipeline from this laser-scanned data remains challenging. The data acquired from the plant facility can be incomplete due to complex occlusion, or it can be affected by noise due to the reflective surfaces of the pipelines and other parts. The aim of this study is to propose a method for generating models of entire pipelines that include straight pipes, elbows, reducers, and tee pipes from laserscanned data. The proposed 3D reconstruction method for as-built pipelines is divided into three main tasks: (1) identifying the types and locations of the pipelines from the laser-scanned data; (2) segmenting the pipelines into each type of pipe form; and (3) reconstructing the pipelines’ geometry and topology and generating models of them. Field experiments were performed at an operating industrial plant in order to validate the proposed method. The results revealed that the proposed method can indeed contribute to the automation of 3D reconstruction of as-built pipelines.

Prediction of government-owned building energy consumption based on an RReliefF and support vector machine model

AbstractAccurate prediction of the energy consumption of government-owned buildings in the design phase is vital for government agencies, as it enables formulation of the early phases of development of such buildings with a view to reducing their environmental impact. The aim of this study was to identify the variables that are associated with energy consumption in government-owned buildings and to propose a predictive model based on those variables. The proposed approach selects relevant variables using the RReliefF variable selection algorithm. The support vector machine (SVM) method is used to develop a model of energy consumption based on the identified variables. The proposed approach was analyzed and validated on data for 175 government-owned buildings derived from the 2003 Commercial Building Energy Consumption Survey (CBECS) database. The experimental results revealed that the proposed model is able to predict the energy consumption of government-owned buildings in the design phase with a reasonab...

Data Mining-Based Predictive Model to Determine Project Financial Success Using Project Definition Parameters

The planning stage is important for project development because the majority of important decisions are made at this stage. Having a well-defined project plan will reduce project uncertainty and increase the likelihood of the project’s success. In other words, based on the level of project definition in the planning stage, project success or failure can be predicted. The aim of this study is to generate a predictive model that will forecast project performance in terms of cost, depending on the project definition level during the early stages of the project before a detailed design is started. The predictive model for this study was generated by support vector machine (SVM). A survey of 77 completed construction projects in Korea was conducted in order to collect the project defined level and cost data from each of those projects by questioning selected clients, architects, and construction managers who had participated before beginning the detailed design stage in the project. It is anticipated that prediction results will help clients and project managers revise their project planning when they encounter a poor performance prediction. Furthermore, the research result imply that employing the proposed model can help project participants achieve success by managing projects more effectively.

A Comparative Study on Color Model-Based Concrete Image Retrieval in Different Invariant Color Spaces

Construction progress monitoring has been recognized as one of the key elements that lead to the success of a construction project. The first requirement for effective progress monitoring is the collection and analysis of construction progress information. Through the use of image retrieval, progress information about structural components can be derived from the construction site image. In this paper, the method of color model-based, concrete image retrieval is proposed for utilization in construction progress monitoring. For effective concrete image retrieval, a comparison of concrete color models in four invariant color spaces, such as normalized rgb, HSI, YCbCr, and CIELUV, is conducted. Then, the best color configuration and color space to model the inherent concrete color and to efficiently discriminate between concrete and other objects (or “non-concrete” objects) are determined, using Mahalanobis distance and performance measures. Experimental results show that L-U color configuration in CIELUV color space yield the optimal retrieving performance, and subsequently, the highest retrieval rate of concrete color.