• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.12
• /
• pp.150-157
• /
• 2015

Pedestrian detection has largely been researched as one of the important technologies for autonomous driving vehicle and preventing accidents. There are two categories for pedestrian detection, camera-based and LIDAR-based. LIDAR-based methods have the advantage of the wide angle of view and insensitivity of illuminance change while camera-based methods have not. However, there are several problems with 3D LIDAR, such as insufficient resolution to detect distant pedestrians and decrease in detection rate in a complex situation due to segmentation error and occlusion. In this paper, two methods using GM-PHD filter are proposed to improve the poor rates of pedestrian detection algorithms based on 3D LIDAR. First one improves detection performance and resolution of object by automatic accumulation of points in previous frames onto current objects. Second one additionally enhances the detection results by applying the GM-PHD filter which is modified in order to handle the poor situation to classified multi target. A quantitative evaluation with autonomously acquired road environment data shows the proposed methods highly increase the performance of existing pedestrian detection algorithms.

• Journal of the Korean Society of Safety
• /
• v.36 no.1
• /
• pp.18-25
• /
• 2021

In many developed countries, such as South Korea, efficiently maintaining the aging infrastructures is an important issue. Currently, inspectors visually inspect the infrastructure for maintenance needs, but this method is inefficient due to its high costs, long logistic times, and hazards to the inspectors. Thus, in this paper, a novel crack inspection approach for concrete bridges is proposed using integrated image processing techniques. The proposed approach consists of four steps: (1) training a deep learning model to automatically detect cracks on concrete bridges, (2) acquiring in-situ images using a drone, (3) generating orthomosaic images based on 3D modeling, and (4) detecting cracks on the orthmosaic image using the trained deep learning model. Cascade Mask R-CNN, a state-of-the-art instance segmentation deep learning model, was trained with 3235 crack images that included 2415 hard negative images. We selected the Tancheon overpass, located in Seoul, South Korea, as a testbed for the proposed approach, and we captured images of pier 34-37 and slab 34-36 using a commercial drone. Agisoft Metashape was utilized as a 3D model generation program to generate an orthomosaic of the captured images. We applied the proposed approach to four orthomosaic images that displayed the front, back, left, and right sides of pier 37. Using pixel-level precision referencing visual inspection of the captured images, we evaluated the trained Cascade Mask R-CNN's crack detection performance. At the coping of the front side of pier 37, the model obtained its best precision: 94.34%. It achieved an average precision of 72.93% for the orthomosaics of the four sides of the pier. The test results show that this proposed approach for crack detection can be a suitable alternative to the conventional visual inspection method.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.1 s.307
• /
• pp.45-52
• /
• 2006

In this paper, a real-time stereoscopic image conversion method using a single frame from a 2-D image is proposed. The Stereoscopic image is generated by creating depth map using vortical position information and parallax processing. For a real-time processing of stereoscopic conversion and reduction of hardware complexity, it uses image sampling, object segmentation by standardizing luminance and depth map generation by boundary scan. The proposed method offers realistic 3-D effect regardless of the direction, velocity and scene conversion of the 2-D image. It offers effective stereoscopic conversion using images suitable conditions assumed in this paper such as recorded image at long distance, landscape and panorama photo because it creates different depth sense using vertical position information from a single frame. The proposed method can be applied to still image because it uses a single frame from a 2-D image. The proposed method has been evaluated using visual test and APD for comparing the stereoscopic image of the proposed method with that of MTD. It is confirmed that stereoscopic images conversed by the proposed method offers 3-D effect regardless of the direction and velocity of the 2-D image.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.43 no.6 s.312
• /
• pp.28-35
• /
• 2006

Statistical models of shape variability based on active shape models (ASMs) have been successfully utilized to perform segmentation and recognition tasks in two-dimensional (2D) images. Three-dimensional (3D) model-based approaches are more promising than 2D approaches since they can bring in more realistic shape constraints for recognizing and delineating the object boundary. For 3D model-based approaches, however, building the 3D shape model from a training set of segmented instances of an object is a major challenge and currently it remains an open problem in building the 3D shape model, one essential step is to generate a point distribution model (PDM). Corresponding landmarks must be selected in all1 training shapes for generating PDM, and manual determination of landmark correspondences is very time-consuming, tedious, and error-prone. In this paper, we propose a novel automatic method for generating 3D statistical shape models. Given a set of training 3D shapes, we generate a 3D model by 1) building the mean shape fro]n the distance transform of the training shapes, 2) utilizing a tetrahedron method for automatically selecting landmarks on the mean shape, and 3) subsequently propagating these landmarks to each training shape via a distance labeling method. In this paper, we investigate the accuracy and compactness of the 3D model for the human liver built from 50 segmented individual CT data sets. The proposed method is very general without such assumptions and can be applied to other data sets.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.42 no.5 s.305
• /
• pp.51-58
• /
• 2005

This paper proposes an efficient algorithm for extracting a tumor region from an breast ultrasound volume image by using rate-distortion (R-D) based seeded region growing. In the proposed algorithm the rate and the distortion represent the roughness of the contour and the dissimilarity of pixels in a region, respectively. Staring from an initial seed region set in each cutting plane of a volume, a pair of the seed region and one of adjacent regions whose R-D cost is minimal is searched and then they are merged into a new updated seed region. This procedure is recursively performed until the averaged R-D cost values per the number of contour pixels in the seed region becomes maxim. As a result, the final seed region has good pixel homogeneity and a much smooth contour. Finally, the tumor volume is extracted using the contours of the final seed regions in all the cutting planes. Experimental results show that the averaged error rate of the proposed method is shown to be below 4%.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.17 no.3
• /
• pp.288-296
• /
• 1992

This paper describes an improved 2D QPS(quadrature polar separable) filter design and its applications to texture processing. The filter kernel pair consists of the product of a radial weighting function based on the finite PSS (prolate spheroidal sequences) and an exponential at tenuation function for the orientational angle. It is quadrature and polar separable in the frequency domain. It is near optimal in the energy loss because we let the orientational angle function approximate the radial weighting function. The filter frequency characteristics is easy to control as it depends only upon the design specifications such as the bandwidth, the directional angle, the attenuation constant, and the shift constant of the central frequency. Some applications of the filter in texture processing, such as the generation of the texture image, the estimation of orientation angles, and the segmentations for the synthetic texture image, are considered. The result shows that the filter with the wide bandwidth can be used for the generation of discrimination of the strong orientational textures and the segmentation results are good.

• Journal of the Institute of Convergence Signal Processing
• /
• v.12 no.4
• /
• pp.267-273
• /
• 2011

In this paper, we proposed the method to detect brain tumor region in MR images. Our method is composed of 3 parts, detection of tumor slice, detection of tumor region and tumor boundary detection. In the tumor slice detection step, a slice which contains tumor regions is distinguished using symmetric analysis in 3D brain volume. The tumor region detection step is the process to segment the tumor region in the slice distinguished as a tumor slice. And tumor region is finally detected, using spatial feature and symmetric analysis based on the cluster information. The process for detecting tumor slice and tumor region have advantages which are robust for noise and requires less computational time, using the knowledge of the brain tumor and cluster-based on symmetric analysis. And we use the level set method with fast marching algorithm to detect the tumor boundary. It is performed to find the tumor boundary for all other slices using the initial seeds derived from the previous or later slice until the tumor region is vanished. It requires less computational time because every procedure is not performed for all slices.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1485-1490
• /
• 2004

Recently, we have remarkable developments in intelligent robot systems. The remarkable features of intelligent robot are that it can track user, does face recognition and vital for many surveillance based systems. Advantage of face recognition when compared with other biometrics recognition is that coerciveness and contact that usually exist when we acquire characteristics do not exist in face recognition. However, the accuracy of face recognition is lower than other biometric recognition due to decrease in dimension from of image acquisition step and various changes associated with face pose and background. Factors that deteriorate performance of face recognition are many such as distance from camera to face, lighting change, pose change, and change of facial expression. In this paper, we implement a new 3D active camera system to prevent various pose variation that influence face recognition performance and propose face recognition algorithm for intelligent surveillance system and mobile robot system.

• Journal of Broadcast Engineering
• /
• v.21 no.1
• /
• pp.24-35
• /
• 2016

By estimating environment light distribution, we can generate realistic shadow images in AR(augmented reality). When we estimate light distribution without sensing equipment, environment light model, geometry of virtual object, and surface reflection property are needed. Previous study using 3D marker builds surrounding light environment with a geodesic dome model and analyzes shadow images. Because this method employs candidate shadow maps in initial scene setup, however, it is difficult to estimate precise light information. This paper presents a novel light estimation method based on hierarchical light distribution model subdivision. By using an overlapping area ratio of the segmented shadow and candidate shadow map, we can make hierarchical subdivision of light geodesic dome.

• 제어로봇시스템학회:학술대회논문집
• /
• 1988.10b
• /
• pp.826-831
• /
• 1988

This paper presents an approach to segment an image into areas of surfaces, and to compute the surface properties from a gray-scale image in order to describe the surfaces for reconstruction of the 3-D shape of the objects. In general, an rigid body has several surfaces and many edges. But if it is not polyhedoron, it is necessary not only to describe the relation between surfaces, i.e. its line drawings but also to represent the surfaces' equations itself. In order to compute the surfaces' equation we use a probability of edge distribution. At first it is extracted edges from a gray-level image as much as possible. These are not only the points that maximize the change of an image intensuty but candidates which can be seemed to be edges. Next, other character of a surface (color, coordinates and image intensity) are extracted. In our study, we call the all feature of a surface as "texture", for example color, intensity level, orientation of an edge, shape of a surface and so on. These features of a surface on a pixel of an image plane are mapped to a point of the feature space, and segmented to each groups by cluster analysis on this space. These groups are considered to represent object surface in an image plane. Finally, the states of object surface in 3-D space are computed from distributional probability of local and overall statistical features of a surface, and from shape of a surface.a surface.