Deep Learning on Construction Sites: A Case Study of Sparse Data Learning Techniques for Rebar Segmentation

With recent advancements in deep learning models for image interpretation, it has finally become possible to automate construction site monitoring processes that rely on remote sensing. However, the major drawback of these models is their dependency on large datasets of training images labeled at pixel level, which have to be produced manually by skilled personnel. To alleviate the need for training data, this study evaluates weakly- and semi-supervised semantic segmentation models for construction site imagery to efficiently automate monitoring tasks. As a case study, we compare fully-, weakly- and semi-supervised methods for the detection of rebar covers, which are useful for quality control. In the experiments, recent models, i.e. IRNet, DeepLabv3+ and the cross-consistency training model, are compared for their ability to segment rebar covers from construction site imagery with minimal manual input. The results show that weakly- and semi-supervised models can indeed approach the performance of fully-supervised models, with the majority of the target objects being properly found. Through this study, construction site stakeholders are provided with detailed information on how tp leverage deep learning for efficient construction site monitoring and weigh preprocessing, training and testing efforts against each other in order to decide between fully-, weakly- and semi-supervised training.

Semantic Analysis of Big Data in the Problem of Forecasting the Innovative Development of the Energy Infrastructure of the Russian Federation

The article discusses the use of methods of semantic analysis and natural language processing to support research and forecasting the innovative development of the energy infrastructure of the Russian Federation. The existing methods and approaches to the organization of monitoring of technological solutions and innovative scientific developments are considered. To automate monitoring, the authors propose the use of natural language processing (NLP) methods. Semantic analysis and knowledge integration are based on a system of ontologies. The paper presents the main methods and approaches to building an infrastructure for processing open Big Data. Application of the proposed methods makes it possible to improve the quality of scientific research in this area and make them better.

Experimental Flight Patterns Evaluation for a UAV-Based Air Pollutant Sensor

The use of drones in combination with remote sensors have displayed increasing interest over the last years due to its potential to automate monitoring processes. In this study, a novel approach of a small flying e-nose is proposed by assembling a set of AlphaSense electrochemical-sensors to a DJI Matrix 100 unmanned aerial vehicle (UAV). The system was tested on an outdoor field with a source of NO2. Field tests were conducted in a 100 m2 area on two dates with different wind speed levels varying from low (0.0–2.9m/s) to high (2.1–5.3m/s), two flight patterns zigzag and spiral and at three altitudes (3, 6 and 9 m). The objective of this study is to evaluate the sensors responsiveness and performance when subject to distinct flying conditions. A Wilcoxon rank-sum test showed significant difference between flight patterns only under High Wind conditions, with Spiral flights being slightly superior than Zigzag. With the aim of contributing to other studies in the same field, the data used in this analysis will be shared with the scientific community.

PROPOSAL OF A VISION SYSTEM FOR AUTOMATED LINE MPS 500

Urgency of the research. Research needs from this area are based on designing of effective and affordable vision system solution with aim of automation level increasing in content industry 4.0 and should be an advantageous solution mainly for SMEs. In overall, development direction in vision system area pointed to the necessity for innovative technologies implementation that starts from supply chains up to customers. Target setting. Main aim of article is to propose a solution for image processing of selected assembly parts at specified automated line from FESTO company to automate monitoring and evaluating of obtained data together with supporting of educational activities for field: „automation and control of machines and processes“ of our students, at other hand. Actual scientific researches and issues analysis. Currently, vision systems have enjoyed a great popularity, their implementation into the automated lines grown up and application range more and wider. Supporting from manufacturers is strong, so far, we consider that useful and well-priced solution will be benefit in research area. Uninvestigated parts of general matters defining. Existing realized and implemented solutions are based on solid whole concept from suppliers, starting from connectivity through image processing up to evaluation of situations (pictures, state etc...). The research objective. Purpose of an article is to provide an alternative proposal for implementation of affordable and alternative vision system solution into the selected automated line from FESTO company. The statement of basic materials. Most of realized projects are based on complex vision system solution. Customized and well – priced proposal are rarely, so, we consider, that should be useful to contribute into research community in form of an article from this area. Conclusions. Presented article offer fundamental deployment of vision system into the automated line from company FESTO with aim of intelligence level increasing of this line. Last, but not least, purpose of automated line will be educational training with focus to experimental verification of students knowledges, primarily from pneumatics, compact PLC and vision system, of course.

Automated Building Process Monitoring

Monitoring of building process activities is the basis for effective control and management of a building project. In its traditional way it is, however, time consuming, inaccurate and expensive. To improve the monitoring process researchers are investigating methods to automate monitoring and support project managers with accurate and timely information about activity progress. The chapter describes some of these methods and then concentrates on a solution, which takes into account all three aspects of project management: coordination, control and communication. Activity progress is monitored directly by using a combination of data collection methods, which are based on the building information model (BIM), especially on the 4D model of the building. The resulting system is described, evaluated and discussed.