• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.6
• /
• pp.175-182
• /
• 2020

The non-contact eye tracking is a nonintrusive human-computer interface providing hands-free communications for people with severe disabilities. Recently. it is expected to do an important role in non-contact systems due to the recent coronavirus COVID-19, etc. This paper proposes a novel approach for an eye mouse using an eye tracking method based on a context-aware based AdaBoost multi-region classifier and ASSL algorithm. The conventional AdaBoost algorithm, however, cannot provide sufficiently reliable performance in face tracking for eye cursor pointing estimation, because it cannot take advantage of the spatial context relations among facial features. Therefore, we propose the eye-region context based AdaBoost multiple classifier for the efficient non-contact gaze tracking and mouse implementation. The proposed method detects, tracks, and aggregates various eye features to evaluate the gaze and adjusts active and semi-supervised learning based on the on-screen cursor. The proposed system has been successfully employed in eye location, and it can also be used to detect and track eye features. This system controls the computer cursor along the user's gaze and it was postprocessing by applying Gaussian modeling to prevent shaking during the real-time tracking using Kalman filter. In this system, target objects were randomly generated and the eye tracking performance was analyzed according to the Fits law in real time. It is expected that the utilization of non-contact interfaces.

• Journal of Digital Convergence
• /
• v.11 no.2
• /
• pp.243-248
• /
• 2013

This paper proposes an effective method using Adaptive Boosting to track a person's head in complex background. By only one way to feature extraction methods are not sufficient for modeling a person's head. Therefore, the method proposed in this paper, several feature extraction methods for the accuracy of the detection head running at the same time. Feature Extraction for the imaging of the head was extracted using sub-region and Haar wavelet transform. Sub-region represents the local characteristics of the head, Haar wavelet transform can indicate the frequency characteristics of face. Therefore, if we use them to extract the features of face, effective modeling is possible. In the proposed method to track down the man's head from the input video in real time, we ues the results after learning Harr-wavelet characteristics of the three types using AdaBoosting algorithm. Originally the AdaBoosting algorithm, there is a very long learning time, if learning data was changes, and then it is need to be performed learning again. In order to overcome this shortcoming, in this research propose efficient method using cascade AdaBoosting. This method reduces the learning time for the imaging of the head, and can respond effectively to changes in the learning data. The proposed method generated classifier with excellent performance using less learning time and learning data. In addition, this method accurately detect and track head of person from a variety of head data in real-time video images.

• Journal of the Korea Society of Computer and Information
• /
• v.3 no.1
• /
• pp.81-94
• /
• 1998

In order to extract the contour of interesting object in the image, Kass suggested the Active Contour Model called "Snakes". Snakes is a model which defines the contour of image energy. It also can find the contour of object by minimizing these energy functions. The speed of this model is slow and this model is sensitive of initialization. In order to improve these problems, Gunn extracted the accurate contour by using two initialization. and operated to less sensitive of initialization. This method could extract more accurate contour than the existing method, but it had no effect in the speed and it was sensitive of noise. This paper applied to the Energy Minimization Algorithm about only the pixel within the window applying the window of 8$\times$8 size at each contour point consisting Snakes in order to solve these problems. The method offered in this paper is applied to extract the contour of original image and cup image added to gaussian noise. By tracking the face using this offered method, it is applied to virtual reality and motion tracking. tracking.

• The Journal of the Korea Contents Association
• /
• v.19 no.6
• /
• pp.105-113
• /
• 2019

This study is to analyze the concentration of visual perception of the subject through the eye-tracking experiments targeting popular characters and to investigate it is effective to highlight which part of the character in inducing consumers' interest and curiosity. As a result of the study, the subjects paid more attention to the face than the character's body, and also noted a little attention to auxiliary elements such as pants, tees, shoes, and whiskers. These results mean there is clearly different factor in which the subject's gaze on the character frequently stays and does for a long time. Therefore, the designers should recognize the differences of the concentration of visual perception by character components as a differentiating factor to feel the emotional satisfaction such as the user's interest, empathy, and immersion in the character and actively reflect it when developing the character design. In addition, there was a difference in the visual perception concentration of each of the auxiliary elements, and if careful research is done on whether the influence factors are also due to individual characteristics or another influential factors, it will be a wider study on character design.

• 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.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.9
• /
• pp.3782-3796
• /
• 2020

A three-dimensional (3D) reconstruction is an important research area in computer vision. The ability to detect and match features across multiple views of a scene is a critical initial step. The tracking matrix W obtained from a 3D reconstruction can be applied to structure from motion (SFM) algorithms for 3D modeling. We often fail to generate an acceptable number of features when processing face or medical images because such images typically contain large homogeneous regions with minimal variation in intensity. In this study, we seek to locate sufficient matching points not only in general images but also in face and medical images, where it is difficult to determine the feature points. The algorithm is implemented on an adaptive threshold value, a scale invariant feature transform (SIFT), affine SIFT, speeded up robust features (SURF), and affine SURF. By applying the algorithm to face and general images and studying the geometric errors, we can achieve quasi-dense matching points that satisfy well-functioning geometric constraints. We also demonstrate a 3D reconstruction with a respectable performance by applying a column space fitting algorithm, which is an SFM algorithm.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.1202-1205
• /
• 2005

This paper describes a method for vision-based person identification that can detect, track, and recognize person from video using multiple cues: height and dressing colors. The method does not require constrained target's pose or fully frontal face image to identify the person. First, the system, which is connected to a pan-tilt-zoom camera, detects target using motion detection and human cardboard model. The system keeps tracking the moving target while it is trying to identify whether it is a human and identify who it is among the registered persons in the database. To segment the moving target from the background scene, we employ a version of background subtraction technique and some spatial filtering. Once the target is segmented, we then align the target with the generic human cardboard model to verify whether the detected target is a human. If the target is identified as a human, the card board model is also used to segment the body parts to obtain some salient features such as head, torso, and legs. The whole body silhouette is also analyzed to obtain the target's shape information such as height and slimness. We then use these multiple cues (at present, we uses shirt color, trousers color, and body height) to recognize the target using a supervised self-organization process. We preliminary tested the system on a set of 5 subjects with multiple clothes. The recognition rate is 100% if the person is wearing the clothes that were learned before. In case a person wears new dresses the system fail to identify. This means height is not enough to classify persons. We plan to extend the work by adding more cues such as skin color, and face recognition by utilizing the zoom capability of the camera to obtain high resolution view of face; then, evaluate the system with more subjects.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.11
• /
• pp.179-186
• /
• 2009

In this paper, we propose a head detection method based on vertical and horizontal pixel histogram analysis in order to overcome drawbacks of the previous head detection approach using Haar-like feature-based face detection. In the proposed method, we create the vertical and horizontal foreground pixel histogram images from the background subtraction image, which represent the number of foreground pixels in the same vertical or horizontal position. Then we extract feature points of a head region by applying Harris corner detection method to the foreground pixel histogram images and by analyzing corner points. The proposal method shows robust head detection results even in the face image covering forelock by hairs or the back view image in which the previous approaches cannot detect the head regions.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.4
• /
• pp.159-166
• /
• 2016

In this paper, an active object tracking system basing on the pan/tilt-embedded stereo camera system is suggested and implemented. In the proposed system, once the face area of a target is detected from the input stereo image by using a YCbCr color model and phase-type correlation scheme and then, using this data as well as the geometric information of the tracking system, the distance and 3D information of the target are effectively extracted in real-time. Basing on these extracted data the pan/tilted-embedded stereo camera system is adaptively controlled and as a result, the proposed system can track the target adaptively under the various circumstance of the target. From some experiments using 480 frames of the test input stereo image, it is analyzed that a standard variation between the measured and computed the estimated target's height and an error ratio between the measured and computed 3D coordinate values of the target is also kept to be very low value of 1.03 and 1.18% on average, respectively. From these good experimental results a possibility of implementing a new real-time intelligent stereo target tracking and surveillance system using the proposed scheme is finally suggested.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.11
• /
• pp.180-187
• /
• 2013

The dinger in solar generation is the amount of sunlight which the solar cells absorb. Various types of solar trackers, which rotate themselves in order to make the solar cells face the sun as much as possible, have been developed, and especially the method of tracking with two axes has greatly contributed in increasing the generation amount at work sites. Among theses 2-axis solar trackers, the inclination-dependent 2-axis solar tracker are widely utilized for its advantages of requiring less initial investment and easy maintenance due to a solid structure. However, the drawback is that the generation efficiency is relatively low because of the structural restriction that limits the rotation angle, thus making it less efficient when tracking the sun. This paper proposes a method to increase the generation efficiency of the inclination-dependent 2-axis solar tracker. It also contains the derived equations needed for precise controlling along with a method to keep tracking with the other axis even when one has reached its angle limit. To confirm that the proposed method increases the amount of incidence onto the solar cells, formulas needed for operation on the proposed method and tracking the exact position of the sun are derived, and applying this to the quarterly data of Korea Astronomy and Space Science Institute it shows maximum over 11.1% more incidence compared to existing methods.