• Annual Conference of KIPS
• /
• 2019.10a
• /
• pp.1002-1003
• /
• 2019

Face recognition is an extensively researched area of computer vision. Visible, infrared, thermal, and 3D modalities have been used against various challenges of face recognition such as illumination, pose, expression, partial information, and disguise. In this paper we present a multi-modal approach to face recognition using convolutional neural networks. We use visible and thermal face images as two separate inputs to a multi-input deep learning network for face recognition. The experiments are performed on IRIS visible and thermal face database and high face verification rates are achieved.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.1
• /
• pp.166-179
• /
• 2021

Recently, with the development of technology, computer vision research based on the human visual system has been actively conducted. Saliency maps have been used to highlight areas that are visually interesting within the image, but they can suffer from low performance due to external factors, such as an indistinct background or light source. In this study, existing color, brightness, and contrast feature maps are subjected to multiple shape and orientation filters and then connected to a fully connected layer to determine pixel intensities within the image based on location-based weights. The proposed method demonstrates better performance in separating the background from the area of interest in terms of color and brightness in the presence of external elements and noise. Location-based weight normalization is also effective in removing pixels with high intensity that are outside of the image or in non-interest regions. Our proposed method also demonstrates that multi-filter normalization can be processed faster using parallel processing.

• Journal of Information Processing Systems
• /
• v.18 no.6
• /
• pp.794-802
• /
• 2022

The recovery of reasonable depth information from different scenes is a popular topic in the field of computer vision. For generating depth maps with better details, we present an efficacious monocular depth prediction framework with coordinate attention and feature fusion. Specifically, the proposed framework contains attention, multi-scale and feature fusion modules. The attention module improves features based on coordinate attention to enhance the predicted effect, whereas the multi-scale module integrates useful low- and high-level contextual features with higher resolution. Moreover, we developed a feature fusion module to combine the heterogeneous features to generate high-quality depth outputs. We also designed a hybrid loss function that measures prediction errors from the perspective of depth and scale-invariant gradients, which contribute to preserving rich details. We conducted the experiments on public RGBD datasets, and the evaluation results show that the proposed scheme can considerably enhance the accuracy of depth prediction, achieving 0.051 for log10 and 0.992 for δ<1.25³ on the NYUv2 dataset.

• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.648-651
• /
• 2023

Video anomaly detection aims to detect abnormal events. Motivated by the power of transformers recently shown in vision tasks, we propose a novel transformer-based network for video anomaly detection. To capture long-range information in video, we employ a multi-scale transformer as an encoder. A convolutional decoder is utilized to predict the future frame from the extracted multi-scale feature maps. The proposed method is evaluated on three benchmark datasets: USCD Ped2, CUHK Avenue, and ShanghaiTech. The results show that the proposed method achieves better performance compared to recent methods.

• Journal of Biosystems Engineering
• /
• v.31 no.1 s.114
• /
• pp.59-64
• /
• 2006

Application of herbicide for rice cropping is inevitable but notorious for its side effect of environmental pollution. Precision fanning will be one of important tools for the least input and sustainable fanning and could be achieved by implementation of the variable rating technology. If a device to detect weeds in rice field is available, herbicide could be applied only to the places where it is needed by the manner of the variable rating technology. The study was carried out to develop an algorithm of image processing to detect weeds in rice field using a machine vision system of multi-spectral digital images. A series of multi-spectral rice field picture of 560, 680 and 800 nm of center wavelengths were acquired from the 27th day to the 39th day after transplanting in the ineffective tillering stage of a rice growing period. A discrimination model to distinguish pixels of weeds from those of rice plant and weed image was developed. The model was proved as having accuracies of 83.6% and 58.9% for identifying the rice plant and the weed, respectively. The model was used in the algorithm to differentiate weed images from mingled images of rice plant and weed in a frame of rice field picture. The developed algorithm was tested with the acquired rice field pictures and resulted that 82.7%, 11.9% and 5.4% of weeds in the pictures were noted as the correctly detected, the undetected and the misclassified as rice, respectively, and 81.9% and 18.0% of rice plants in the pictures were marked as the correctly detected and the misclassified as weed, respectively.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.5
• /
• pp.151-157
• /
• 2011

This study analyzed the posture balance of time variation for exercising body a period of time. Posture balance measured output values for the posture balance system of body moving in the multi-parameter. Posture moving variation had three methods such as open and closed eye, head moving and upper body moving. There were checked a parameter that measured vision, vestibular, somatosensory, CNS. This system was evaluated a data through the stability. This system has catched a signal for physical condition of body data such as a data acquisition system, data signal processing and feedback system. The output signal was generated Fourier analysis that using frequency of 0.1Hz, 0.1-0.5Hz, 0.5-1Hz and 1Hz over. The posture balance system will be used to support assessment for body moving the posture balance of time variation. It was expected to monitor a physical parameter for health verification system.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.47 no.2
• /
• pp.68-76
• /
• 2010

Touch-enabled technology is increasingly being accepted as a main communication interface between human and computers. However, conventional touchscreen technologies, such as resistive overlay, capacitive overlay, and SAW(Surface Acoustic Wave), are not cost-effective for large screens. As an alternative to the conventional methods, we introduce a newly emerging method, an optical imaging touchscreen which is much simpler and more cost-effective. Despite its attractive benefits, optical imaging touchscreen has to overcome some problems, such as heavy computational complexity, intermittent ghost points, and over-sensitivity, to be commercially used. Therefore, we designed a hardware controller for signal processing and multi-coordinate computation, and proposed Infrared-blocked DA(Dark Area) manipulation as a solution. While the entire optical touch control took 34ms with a 32-bit microprocessor, the designed hardware controller can manage 2 valid coordinates at 200fps and also reduce energy consumption of infrared diodes from 1.8Wh to 0.0072Wh.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.10
• /
• pp.19-28
• /
• 2022

In this paper, we propose the Detector Scheduler which determines the best tracking-by-detection (TBD) mechanism to perform real-time high-accurate multi-object tracking (MOT). The Detector Scheduler determines whether to run a detector by measuring the dissimilarity of features between different frames. Furthermore, we propose a self-supervision method to learn the Detector Scheduler with tracking results since it is difficult to generate ground truth (GT) for learning the Detector Scheduler. Our proposed self-supervision method generates pseudo labels on whether to run a detector when the dissimilarity of the object cardinality or appearance between frames increases. To this end, we propose the Detector Scheduling Loss to learn the Detector Scheduler. As a result, our proposed method achieves real-time high-accurate multi-object tracking by boosting the overall tracking speed while keeping the tracking accuracy at most.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.1840-1843
• /
• 1997

The task environment of a robot is changing rapidly and task itself becomes complicated due to current industrial trends of multi-product and small lot size production. A convenient user-interfaced off-line programming(OLP) system is being developed in order to overcome the difficulty in teaching a robot task. Using the OLP system, operators can easily teach robot tasks off-line and verify feasibility of the task through simulation of a robot prior to the on-line execution. However, some task errors are inevitable by kinematic differences between the robot model in OLP and the actual robot. Three calibration methods using image information are proposed to compensate the kinematic differences. These methods compose of a relative position vector method, three point compensation method, and base line compensation method. To compensate a kinematic differences the vision system with one monochrome camera is used in the calibration experiment.

• International Journal of Automotive Technology
• /
• v.8 no.4
• /
• pp.513-520
• /
• 2007

In this paper, we propose an approach to identify the driving environment for intelligent highway vehicles by means of image processing and computer vision techniques. The proposed approach mainly consists of two consecutive computational steps. The first step is the lane marking detection, which is used to identify the location of the host vehicle and road geometry. In this step, related standard image processing techniques are adapted for lane-related information. In the second step, by using the output from the first step, a four-stage algorithm for vehicle detection is proposed to provide information on the relative position and speed between the host vehicle and each preceding vehicle. The proposed approach has been validated in several real-world scenarios. Herein, experimental results indicate low false alarm and low false dismissal and have demonstrated the robustness of the proposed detection approach.