• International conference on construction engineering and project management
• /
• 2022.06a
• /
• pp.877-885
• /
• 2022

Vision-based activity recognition has been widely attempted at construction sites to estimate productivity and enhance workers' health and safety. Previous studies have focused on extracting an individual worker's postural information from sequential image frames for activity recognition. However, various trades of workers perform different tasks with similar postural patterns, which degrades the performance of activity recognition based on postural information. To this end, this research exploited a concept of human-object interaction, the interaction between a worker and their surrounding objects, considering the fact that trade workers interact with a specific object (e.g., working tools or construction materials) relevant to their trades. This research developed an approach to understand the context from sequential image frames based on four features: posture, object, spatial features, and temporal feature. Both posture and object features were used to analyze the interaction between the worker and the target object, and the other two features were used to detect movements from the entire region of image frames in both temporal and spatial domains. The developed approach used convolutional neural networks (CNN) for feature extractors and activity classifiers and long short-term memory (LSTM) was also used as an activity classifier. The developed approach provided an average accuracy of 85.96% for classifying 12 target construction tasks performed by two trades of workers, which was higher than two benchmark models. This experimental result indicated that integrating a concept of the human-object interaction offers great benefits in activity recognition when various trade workers coexist in a scene.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.35 no.2
• /
• pp.103-112
• /
• 2017

Due to the large size and high complexity of modern buildings, the interest and the studies about indoor spatial information are increasing. Previous studies related to indoor spatial information were mostly about relevant technologies, and the application of indoor spatial information has been less studied. In the present study, the public administrative work areas where indoor spatial information may be applied were identified by using a modified delphi technique. And the indoor LOD (Level of Detail) and indoor positioning accuracy for indoor spatial information applications considering user requirements was established as standards for efficiently establishing and providing services. The required LOD and positioning accuracy for services was established by reestablishing indoor LOD and positioning accuracy and classifying services with reference to those. The indoor LOD was reestablished from LOD 0 to 4 by focusing on service utilization and general recognition, and the positioning accuracy was reestablished in three levels by considering the accuracy of the present positioning technology and service utilization status.

• Journal of information and communication convergence engineering
• /
• v.13 no.2
• /
• pp.105-112
• /
• 2015

The estimation of missing sensor values is an important problem in sensor network applications, but the existing approaches have some limitations, such as the limitations of application scope and estimation accuracy. Therefore, in this paper, we propose a new estimation model based on a spatial-temporal correlation analysis (STCAM). STCAM can make full use of spatial and temporal correlations and can recognize whether the sensor parameters have a spatial correlation or a temporal correlation, and whether the missing sensor data are continuous. According to the recognition results, STCAM can choose one of the most suitable algorithms from among linear interpolation algorithm of temporal correlation analysis (TCA-LI), multiple regression algorithm of temporal correlation analysis (TCA-MR), spatial correlation analysis (SCA), spatial-temporal correlation analysis (STCA) to estimate the missing sensor data. STCAM was evaluated over Intel lab dataset and a traffic dataset, and the simulation experiment results show that STCAM has good estimation accuracy.

• Journal of Korean Society for Geospatial Information Science
• /
• v.9 no.2 s.18
• /
• pp.37-46
• /
• 2001

Spectral pattern recognition techniques are most used in classification of remotely sensed data. Yet, in any real image, adjacent pixels are related, because imaging sensors acquire significant portions of energy from adjacent pixels. And, with the continued improvement in the spatial resolution of remote sensing systems, another spatial pattern recognition approach is must considered. In this study, we aim to show the potentiality of spatial classification methods through comparing the accuracies of spectral classification methods and those of spectral classification methods. By the comparisons between the two methods, classification accuracies of 6 different spatial classification methods are higher than that of spectral classification method by 2-6% or so. Additionally, we can show it statistically through the classification experiments with different band combinations.

• KIEAE Journal
• /
• v.11 no.6
• /
• pp.95-100
• /
• 2011

Although Space Syntax is a well-known technique for spatial analysis, debates have taken place among some researchers because the Space Syntax discards geometric information as both shapes and sizes of spaces, and hence may cause some inconsistencies. Therefore, this study aims at developing methods to recognize and manage spatial shapes for more precise space syntax analysis. To reach this goal, this study employed both a graph theory and binary spatial partitioning (BSP) tree to recognize and manage spatial information. As a result, spatial shapes and sizes could be recognized by checking loops in graph converted from spatial shapes of built environment. Each spatial shape could be managed sequentially by BSP tree with hierarchical structure. Through such recognition and management processes, convex maps composed of the fattest and fewest convex spaces could be drawn. In conclusion, we hope that the methods developed here will be useful for urban planning to find appropriate purposes of spaces to satisfy the sustainability of built environment on the basis of the spatial and social relationships in urban spaces.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.15 no.3
• /
• pp.298-305
• /
• 2012

In this paper, we propose an integrated algorithm for scene recognition, which has been a challenging computer vision problem, with application to mobile robot localization. The proposed scene recognition method utilizes SIFT and visual words as local-level features and GIST as a global-level feature. As local-level and global-level features complement each other, it results in improved performance for scene recognition. This improved algorithm is of low computational complexity and robust to image distortions.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.1E
• /
• pp.20-24
• /
• 1994

A new speech recognition model, DLSMM(Dynamically Localized Self-organizing Map Model) and its effective training algorithm are proposed in this paper. In DLSMM, temporal and spatial distortions of speech are efficiently normalized by dynamic programming technique and localized self-organizing maps, respectively. Experiments on Korean digits recognition have been carried out. DLSMM has smaller Experiments on Korean digits recognition have been carried out. DLSMM has smaller connections than predictive neural network models, but it has scored a little high recognition rate.

• Journal of Information Processing Systems
• /
• v.17 no.2
• /
• pp.385-398
• /
• 2021

The complexity of deep learning models affects the real-time performance of gesture recognition, thereby limiting the application of gesture recognition algorithms in actual scenarios. Hence, a residual learning neural network based on a deep convolutional neural network is proposed. First, small convolution kernels are used to extract the local details of gesture images. Subsequently, a shallow residual structure is built to share weights, thereby avoiding gradient disappearance or gradient explosion as the network layer deepens; consequently, the difficulty of model optimisation is simplified. Additional convolutional neural networks are used to accelerate the refinement of deep abstract features based on the spatial importance of the gesture feature distribution. Finally, a fully connected cascade softmax classifier is used to complete the gesture recognition. Compared with the dense connection multiplexing feature information network, the proposed algorithm is optimised in feature multiplexing to avoid performance fluctuations caused by feature redundancy. Experimental results from the ISOGD gesture dataset and Gesture dataset prove that the proposed algorithm affords a fast convergence speed and high accuracy.

• Journal of Korea Spatial Information System Society
• /
• v.1 no.1 s.1
• /
• pp.49-55
• /
• 1999

Extracting information of maps is necessary to establish the GIS. In this paper, a house recognition method on the scanned topographic map is described. A contour detection method is used to extract houses from the scanned maps and RLE (run-length encoding) method is used for manipulating houses touching grid lines. To handle houses touched to roads and borderlines, morphological operation is used. To remove misrecognition occurred by morphological operation, the legions which contain characters on the map are also automatically eliminated.

• Journal of Biomedical Engineering Research
• /
• v.16 no.1
• /
• pp.33-42
• /
• 1995

A new neural network architecture for the recognition of patterns from images is proposed, which is partially based on the results of physiological studies. The proposed network is composed of multi-layers and the nerve cells in each layer are connected by spatial filters which approximate receptive fields in optic nerve fields. In the proposed method, patterns recognition for complicated images is carried out using global features as well as local features such as lines and end-points. A new generating method of matched filters representing global features is proposed in this network.