• Proceedings of the KSRS Conference
• /
• v.2
• /
• pp.610-614
• /
• 2006

In this paper the advanced method of geodesic active contours was developed for the task of building detection from aerial and satellite images. Automatic extraction of man-made structures including buildings, building blocks or roads from remote sensing data is useful for land use mapping, scene understanding, robotic navigation, image retrieval, surveillance, emergency management procedures, cadastral etc. A level set method based on a region-driven segmentation model was implemented with which building boundaries were detected, through this curve propagation technique. The essence of this approach is to optimize the position and the geometric form of the curve by measuring information along that curve, and within the regions that compose the image partition. To this end, one can consider uniform intensities inside objects and the background. Thus, given an initial position of the curve, one can determine global, region-driven functions and provide a statistical description of the inside and outside object area. The calculus of variations and a gradient descent method was used to optimize the variational functional by an iterative steady state process. Experimental results demonstrate the potential of the proposed processing scheme.

• Journal of the Korea Society of Computer and Information
• /
• v.12 no.3
• /
• pp.105-113
• /
• 2007

In this paper, an image segmentation method which is based on edge salience map and region merging is presented. The edge salience map is calculated by combining a texture edge map with a color edge map. The texture edge map is computed over multiple spatial orientations and frequencies by using Gabor filter. A color edge is computed over the H component of the HSI color model. Then the Watershed transformation technique is applied to the edge salience map to and homogeneous regions where the dissimilarity of color and texture distribution is relatively low. The Watershed transformation tends to over-segment images. To merge the over-segmented regions, first of all, morphological operation is applied to the edge salience map to enhance a contrast of it and also to find mark regions. Then the region characteristics, a Gabor texture vector and a mean color, in the segmented regions is defined and regions that have the similar characteristics, are merged. Experimental results have demonstrated the superiority in segmentation results for various images.

• Journal of Institute of Control, Robotics and Systems
• /
• v.14 no.5
• /
• pp.438-443
• /
• 2008

In this paper, segmenting and classifying low contrast defects on flat panel display is one of the key problems for automatic inspection system in practice. Problems become more complicated when the quality of acquired image is degraded by the illumination irregularity. Many algorithms are developed and implemented successfully for the defects segmentation. However, vision algorithms are inherently prone to be dependent on parameters to be set manually. In this paper, one morphological segmentation algorithm is chosen and a technique using frequency domain analysis of input images is developed for automatically selection the morphological parameter. An extensive statistical performance analysis is performed to compare the developed algorithms.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.29 no.3
• /
• pp.207-213
• /
• 2009

This paper reports an effort to develop 3D tooth visualization system from CT sequence images as a part of the non-destructive evaluation suitable for the simulation of endodontics, orthodontics and other dental treatments. We focus on the segmentation and visualization for the individual tooth. In dental CT images teeth are touching the adjacent teeth or surrounded by the alveolar bones with similar intensity. We propose an improved level set method with shape prior to separate a tooth from other teeth as well as the alveolar bones. Reconstructed 3D model of individual tooth based on the segmentation results indicates that our technique is a very conducive tool for tooth visualization, evaluation and diagnosis. Some comparative visualization results validate the non-destructive function of our method.

• Proceedings of the IEEK Conference
• /
• 2007.07a
• /
• pp.221-222
• /
• 2007

This paper presents wavelet-based moving object segmentation using double change detection and background registration. Three successive frame differences for detection change were used in the wavelet domain. The background was constructed with the wavelet coefficients in the lowest frequency subband which are the approximated version of an image. Combining double change detection and background registration, we can obtain an efficient moving object segmentation algorithm.

• Computers and Concrete
• /
• v.7 no.6
• /
• pp.511-522
• /
• 2010

A color based segmentation procedure and a modified signature technique have been applied to the detection and analyses of complicated void shapes in cemented materials. The gray-scale segmentation and available signature methods were found to be inefficient especially for the analyses of complicated void shapes. The applicability of the developed methodology has been demonstrated on artificially prepared cemented materials made of self compacted concrete material. In order to characterize the void shapes in the investigated sample images, two new shape parameters called as coefficients of inclusion and exclusion have been proposed. When compared with the traditional use of the signature method, it was found that the methodology followed herein would better characterize complicated void shapes. The methodology followed in this study may be applied to the analysis of complicated void shapes that are often encountered in other cementitious materials such as clays and rocks.

• 전자공학회논문지 IE
• /
• v.43 no.4
• /
• pp.43-51
• /
• 2006

Segmentation method using RSST has many advantages in extracting of accurate region boundaries and controlling the resolution of segmented result and so on. In this paper, we propose three fast RSST algorithms for image segmentation. In first method, we classify links according to weight size for fast link search. In the second method, very similar links before RSST construction are eliminated. In third method, the links of very small regions which are not important for human eye are eliminated. As a result, the total times elapsed for segmentation are reduced by about 10 $\sim$ 40 times, and reconstructed images based on the segmentation results show little degradation of PSNR and visual quality.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.4
• /
• pp.1-8
• /
• 2023

In this paper, we propose a black ice detection platform framework using Convolutional Neural Networks (CNNs). To overcome black ice problem, we introduce a real-time based early warning platform using CNN-based architecture, and furthermore, in order to enhance the accuracy of black ice detection, we apply a multi-scale dilation convolution feature fusion (MsDC-FF) technique. Then, we establish a specialized experimental platform by using a comprehensive dataset of thermal road black ice images for a training and evaluation purpose. Experimental results of a real-time black ice detection platform show the better performance of our proposed network model compared to conventional image segmentation models. Our proposed platform have achieved real-time segmentation of road black ice areas by deploying a road black ice area segmentation network on the edge device Jetson Nano devices. This approach in parallel using multi-scale dilated convolutions with different dilation rates had faster segmentation speeds due to its smaller model parameters. The proposed MsCD-FF Net(2) model had the fastest segmentation speed at 5.53 frame per second (FPS). Thereby encouraging safe driving for motorists and providing decision support for road surface management in the road traffic monitoring department.

• Annual Conference of KIPS
• /
• 2016.10a
• /
• pp.733-734
• /
• 2016

This study proposes a novel approach for ground segmentation of 3D point cloud. We combine two techniques: gradient threshold segmentation, and mean height evaluation. Acquired 3D point cloud is represented as a graph data structures by exploiting the structure of 2D reference image. The ground parts nearing the position of the sensor are segmented based on gradient threshold technique. For sparse regions, we separate the ground and nonground by using a technique called mean height evaluation. The main contribution of this study is a new ground segmentation algorithm which works well with 3D point clouds from various environments. The processing time is acceptable and it allows the algorithm running in real time.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.3 no.3
• /
• pp.8-13
• /
• 2010

Shape form focus (SFF) is a technique that reconstructs 3D shape of an object using image focus. Although many SFF methods have been proposed, there are still notable inaccuracy effects due to noise and non-optimization of image characteristics. In this paper, we propose a noise filter technique for noise reduction and genetic algorithm (GA) for focus measure optimization. The proposed method is analyzed with a statistical criteria such as Root Mean Square Error (RMSE) and correlation.