• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.335-349
• /
• 2023

Bridges constantly undergo deterioration and damage, the most common ones being concrete damage and exposed rebar. Periodic inspection of bridges to identify damages can aid in their quick remediation. Likewise, identifying components can provide context for damage assessment and help gauge a bridge's state of interaction with its surroundings. Current inspection techniques rely on manual site visits, which can be time-consuming and costly. More recently, robotic inspection assisted by autonomous data analytics based on Computer Vision (CV) and Artificial Intelligence (AI) has been viewed as a suitable alternative to manual inspection because of its efficiency and accuracy. To aid research in this avenue, this study performs a comparative assessment of different architectures, loss functions, and ensembling strategies for the autonomous segmentation of bridge components and damages. The experiments lead to several interesting discoveries. Nested Reg-UNet architecture is found to outperform five other state-of-the-art architectures in both damage and component segmentation tasks. The architecture is built by combining a Nested UNet style dense configuration with a pretrained RegNet encoder. In terms of the mean Intersection over Union (mIoU) metric, the Nested Reg-UNet architecture provides an improvement of 2.86% on the damage segmentation task and 1.66% on the component segmentation task compared to the state-of-the-art UNet architecture. Furthermore, it is demonstrated that incorporating the Lovasz-Softmax loss function to counter class imbalance can boost performance by 3.44% in the component segmentation task over the most employed alternative, weighted Cross Entropy (wCE). Finally, weighted softmax ensembling is found to be quite effective when used synchronously with the Nested Reg-UNet architecture by providing mIoU improvement of 0.74% in the component segmentation task and 1.14% in the damage segmentation task over a single-architecture baseline. Overall, the best mIoU of 92.50% for the component segmentation task and 84.19% for the damage segmentation task validate the feasibility of these techniques for autonomous bridge component and damage segmentation using RGB images.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.11
• /
• pp.1720-1728
• /
• 2022

The industrial 4.0 era has been opened with the development of artificial intelligence technology, and the realization of smart farms incorporating ICT technology is receiving great attention in the livestock industry. Among them, the quality management technology of livestock products and livestock operations incorporating computer vision-based artificial intelligence technology represent key technologies. However, the insufficient number of livestock image data for artificial intelligence model training and the severely unbalanced ratio of labels for recognizing a specific defective state are major obstacles to the related research and technology development. To overcome these problems, in this study, combining oversampling and adversarial case generation techniques is proposed as a method necessary to effectively utilizing small data labels for successful defect detection. In addition, experiments comparing performance and time cost of the applicable techniques were conducted. Through experiments, we confirm the validity of the proposed methods and draw utilization strategies from the study results.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.663-668
• /
• 2003

This paper presents a new observation system which is useful to observe the scene of the remote controlled robot vision. This system is composed of a motionless camera and head motion detector with a motion sensor. The motionless camera has a fish eye lens and is for observing a hemisphere space. The head motion detector has a motion sensor is for defining an arbitrary subspace of the hemisphere space from fish eye lens. Thus processing the angular information from the motion sensor appropriately, the direction of face is estimated. However, since the fisheye image is distorted, it is unclear image. The partial domain of a fish eye image is selected by head motion, and this is converted to perspective image. However, since this conversion enlarges the original image spatially and is based on discrete data, crevice is generated in the converted image. To solve this problem, interpolation based on an intensity of the image is performed for the crevice in the converted image (space problem). This paper provides the experimental results of the proposed observation system with the head motion detector and perspective image conversion using the proposed conversion and interpolation methods, and the adequacy and improving point of the proposed techniques are discussed.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.22 no.12
• /
• pp.1084-1088
• /
• 2009

The measurement of dynamic stagger in electric railways is one of the key test parameters to increase speed and maintain safety in electric railways. This paper is introduces a non-contact optical-based measuring instrument of a catenary system in electric railways. The instrument is implemented by utilizing a CCD (Charge Coupled Device) camera installed on the roof of a vehicle for vision acquisition and image processing techniques including the Canny edge detector and the Hough transform to detect contact wires and calculate dynamic stagger. To check the validity of our approach for the intended application, we measured stagger of a overhead wire of a Korea Tilting Train (TTX). The non-contact optical-based measurement system proposed in this paper performs real-time stagger measurement of an activated high-voltage contact wire. By results of this paper, the instrument should be applied to assess performance and reliability of newly developed electric railway vehicles.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2001.09a
• /
• pp.23-27
• /
• 2001

Making the safety situation from the various work system is very important in the industrial fields. The proposed neural network technique is the real titre computation method based theory of inter-node diffusion for searching the safety distances from the sudden appearance-objests during the work driving. The main steps of the distance computation using the theory of stereo vision like the eyes of man is following steps. One is the processing for finding the corresponding points of stereo images and the other is the interpolation processing of full image data from nonlinear image data of obejects. All of them request much memory space and titre. Therefore the most reliable neural-network algorithm is drived for real time recognition of obejects, which is composed of a dynamic programming algorithm based on sequence matching techniques. And the real time reconstruction of nonlinear image information is processed through several simulations. I-D LIPN hardware has been composed, and the real time reconstruction is verified through the various experiments.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.8 no.2
• /
• pp.31-40
• /
• 2012

Face recognition refers to the process of identifying individuals based on their facial features. It has recently become one of the most popular research areas in the fields of computer vision, machine learning, and pattern recognition because it spans numerous consumer applications, such as access control, surveillance, security, credit-card verification, and criminal identification. However, illumination variation on face generally cause performance degradation of face recognition systems under practical environments. Thus, this paper proposes an novel face recognition system using a fusion approach based on local binary pattern and two-dimensional principal component analysis. To minimize illumination effects, the face image undergoes the local binary pattern operation, and the resultant image are divided into two sub-images. Then, two-dimensional principal component analysis algorithm is separately applied to each sub-images. The individual scores obtained from two sub-images are integrated using a weighted-summation rule, and the fused-score is utilized to classify the unknown user. The performance evaluation of the proposed system was performed using the Yale B database and CMU-PIE database, and the proposed method shows the better recognition results in comparison with existing face recognition techniques.

• Journal of IKEEE
• /
• v.25 no.1
• /
• pp.83-87
• /
• 2021

In harsh environments such as fog or fine dust, the cameras' detection ability for object recognition may significantly decrease. In order to accurately obtain important information even in bad weather, fog removal algorithms are necessarily required. Research has been conducted in various ways, such as computer vision/data-based fog removal technology. In those techniques, estimating the amount of fog through the input image's depth information is an important procedure. In this paper, a linear model is presented under the assumption that the image dark channel dictionary, saturation ∗ value, and sharpness characteristics are linearly related to depth information. The proposed method of haze removal through a linear model shows the superiority of algorithm performance in quantitative numerical evaluation.

• International Journal of Internet, Broadcasting and Communication
• /
• v.14 no.4
• /
• pp.58-63
• /
• 2022

Recently, the improvement of computational processing ability due to the rapid development of computing technology has greatly advanced the field of artificial intelligence, and research to apply it in various domains is active. In particular, in the national defense field, attention is paid to intelligent recognition among machine learning techniques, and efforts are being made to develop object identification and monitoring systems using artificial intelligence. To this end, various image processing technologies and object identification algorithms are applied to create a model that can identify friendly and enemy weapon systems and personnel in real-time. In this paper, we conducted image processing and object identification focused on tanks among various weapon systems. We initially conducted processing the tanks' image using a convolutional neural network, a deep learning technique. The feature map was examined and the important characteristics of the tanks crucial for learning were derived. Then, using YOLOv5 Network, a CNN-based object detection network, a model trained by labeling the entire tank and a model trained by labeling only the turret of the tank were created and the results were compared. The model and labeling technique we proposed in this paper can more accurately identify the type of tank and contribute to the intelligent recognition system to be developed in the future.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.19 no.3
• /
• pp.213-220
• /
• 2023

Face recognition is a very important process in video monitoring and is a type of biometric technology. It is mainly used for identification and security purposes, such as ID cards, licenses, and passports. The recognition process has many variables and is complex, so development has been slow. In this paper, we proposed a face recognition method using CNN, which has been re-examined due to the recent development of computers and algorithms, and compared with the feature comparison method, which is an existing face recognition algorithm, to verify performance. The proposed face search method is divided into a face region extraction step and a learning step. For learning, face images were standardized to 50×50 pixels, and learning was conducted while minimizing unnecessary nodes. In this paper, convolution and polling-based techniques, which are one of the deep learning technologies, were used for learning, and 1,000 face images were randomly selected from among 7,000 images of Caltech, and as a result of inspection, the final recognition rate was 98%.

• Korean Journal of Radiology
• /
• v.24 no.7
• /
• pp.698-714
• /
• 2023

In recent years, artificial intelligence, especially object detection-based deep learning in computer vision, has made significant advancements, driven by the development of computing power and the widespread use of graphic processor units. Object detection-based deep learning techniques have been applied in various fields, including the medical imaging domain, where remarkable achievements have been reported in disease detection. However, the application of deep learning does not always guarantee satisfactory performance, and researchers have been employing trial-and-error to identify the factors contributing to performance degradation and enhance their models. Moreover, due to the black-box problem, the intermediate processes of a deep learning network cannot be comprehended by humans; as a result, identifying problems in a deep learning model that exhibits poor performance can be challenging. This article highlights potential issues that may cause performance degradation at each deep learning step in the medical imaging domain and discusses factors that must be considered to improve the performance of deep learning models. Researchers who wish to begin deep learning research can reduce the required amount of trial-and-error by understanding the issues discussed in this study.