• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1997.06a
• /
• pp.131-136
• /
• 1997

We present a method to estimate a dense and sharp depth map using multiple cameras for the application to flexible video production. A key issue for obtaining sharp depth map is how to overcome the harmful influence of occlusion. Thus, we first propose to selectively use the depth information from multiple cameras. With a simple sort and discard technique, we resolve the occlusion problem considerably at a slight sacrifice of noise tolerance. However, boundary overreach of more textured area to less textured area at object boundaries still remains to be solved. We observed that the amount of boundary overreach is less than half the size of the matching window and, unlike usual stereo matching, the boundary overreach with the proposed occlusion-overcoming method shows very abrupt transition. Based on these observations, we propose a hierarchical estimation scheme that attempts to reduce boundary overreach such that edges of the depth map coincide with object boundaries on the one hand, and to reduce noisy estimates due to insufficient size of matching window on the other hand. We show the hierarchical method can produce a sharp depth map for a variety of images.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.4 no.1
• /
• pp.25-31
• /
• 2008

In this paper, we propose a new method that separates background and foreground from stereo images. This method can be improved automatic target tracking system by using disparity map of the stereo vision system and background-separating mask, which can be obtained camera configuration parameters. We use disparity map and camera configuration parameters to separate object from background. Disparity map is made with block matching algorithm from stereo images. A morphology filter is used to compensate disparity error that can be caused by occlusion area. We could obtain a separated object from background when the proposed method was applied to real stereo cameras system.

• Journal of the HCI Society of Korea
• /
• v.1 no.1
• /
• pp.45-52
• /
• 2006

This paper presents an image-space algorithm to real-time collision detection, which is run completely by GPU. For a single object or for multiple objects with no collision, the front and back faces appear alternately along the view direction. However, such alternation is violated when objects collide. Based on these observations, the algorithm propose the depth peeling method which renders the minimal surface of objects, not whole surface, to find colliding. The Depth peeling method utilizes the state-of-the-art functionalities of GPU such as framebuffer object, vertexbuffer object, and occlusion query. Combining these functions, multi-pass rendering and context switch can be done with low overhead. Therefore proposed approach has less rendering times and rendering overhead than previous image-space collision detection. The algorithm can handle deformable objects and complex objects, and its precision is governed by the resolution of the render-target-texture. The experimental results show the feasibility of GPU-based collision detection and its performance gain in real-time applications such as 3D games.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.20 no.9
• /
• pp.49-55
• /
• 2019

This paper presents a marker image evaluation method based on analysis of object distribution in images and classification of images with repetitive patterns for effective marker-based augmented reality (AR) system development. We measure the variance of feature point coordinates to distinguish marker images that are vulnerable to occlusion, since object distribution affects object tracking performance according to partial occlusion in the images. Moreover, we propose a method to classify images suitable for object recognition and tracking based on the fact that the distributions of descriptor vectors among general images and repetitive-pattern images are significantly different. Comprehensive experiments for marker images confirm that the proposed marker image evaluation method distinguishes images vulnerable to occlusion and repetitive-pattern images very well. Furthermore, we suggest that scale-invariant feature transform (SIFT) is superior to speeded up robust features (SURF) in terms of object tracking in marker images. The proposed method provides users with suitability information for various images, and it helps AR systems to be realized more effectively.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.12 no.1
• /
• pp.60-65
• /
• 2012

This paper presents a method for occluded object based motion estimation and tracking system in dynamic image sequences using particle filter with 3D reconstruction. A unique characteristic of this study is its ability to cope with partial occlusion based continuous motion estimation using particle filter inspired from the mirror neuron system in human brain. To update a prior knowledge about the shape or motion of objects, firstly, fundamental 3D reconstruction based occlusion tracing method is applied and object landmarks are determined. And optical flow based motion vector is estimated from the movement of the landmarks. When arbitrary partial occlusions are occurred, the continuous motion of the hidden parts of object can be estimated by particle filter with optical flow. The resistance of the resulting estimation to partial occlusions enables the more accurate detection and handling of more severe occlusions.

• Smart Media Journal
• /
• v.5 no.1
• /
• pp.88-94
• /
• 2016

In this paper, we proposed a new object tracking algorithm based on sparse representation to handle the drifting problem. In APG-L1(accelerated proximal gradient) tracking, the sparse representation is applied to model the appearance of object using linear combination of target templates and trivial templates with proper coefficients. Also, the particle filter based on affine transformation matrix is applied to find the location of object and APG method is used to minimize the l1-norm of sparse representation. In this paper, we make use of the trivial template coefficients actively to block the drifting problem. We experiment the various videos with diverse challenges and the result shows better performance than others.

• Journal of the Korea Computer Graphics Society
• /
• v.21 no.4
• /
• pp.11-17
• /
• 2015

As the needs of real-time visual object tracking are increasing in various kinds of application fields such as surveillance, entertainment, etc., kernel-based mean-shift tracking has received more interests. One of major issues in kernel-based mean-shift tracking is to be robust under partial or full occlusion status. This paper presents a real-time mean-shift tracking which is robust in partial occlusion by applying multi-mode local kernel weight. In the proposed method, a kernel is divided into multiple sub-kernels and each sub-kernel has a kernel weight to be determined according to the location of the sub-kernel. The experimental results show that the proposed method is more stable than the previous methods with multi-mode kernels in partial occlusion circumstance.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• 제어로봇시스템학회:학술대회논문집
• /
• 1990.10b
• /
• pp.917-922
• /
• 1990

Pattern classification is an essential step in automatic robotic assembly which joins together finite number of seperated industrial parts. In this paper, a fast and systematic algorithm for classifying occlusion-free objects is proposed, using the notion of incremental circle transform which describes the boundary contour of an object as a parametric vector function of incremental elements. With similarity transform and line integral, normalized determinant curve of an object classifies each object, independent of position, orientation, scaling of an object and cyclic shift of the stating point for the boundary description.

• Journal of Advanced Navigation Technology
• /
• v.2 no.2
• /
• pp.107-115
• /
• 1998

In this paper, we propose a robust moving object tracking(MOT) method based on motion energy and prediction. MOT consists of two steps: moving object extraction step(MOES) and moving object tracking step(MOTS). For MOES, we use improved motion energy method. For MOTS, we predict the next location of moving object based on distance and direction information among previous instances, so that we can reduce the search space for correspondence. We apply the method to both synthetic and real world sequences and find that the method works well even in the presence of occlusion and disocclusion.