• 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.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.7
• /
• pp.57-64
• /
• 2020

In this paper, we propose an scalable re-detection for correlation filter in visual tracking. In real world, there are lots of target disappearances and reappearances during tracking, thus failure detection and re-detection methods are needed. One of the important point for re-detection is that a searching area must be large enough to find the missing target. For robust visual tracking, we adopt kernelized correlation filter as a baseline. Correlation filters have been extensively studied for visual object tracking in recent years. However conventional correlation filters detect the target in the same size area with the trained filter which is only 2 to 3 times larger than the target. When the target is disappeared for a long time, we need to search a wide area to re-detect the target. Proposed algorithm can search the target in a scalable area, hence the searching area is expanded by 2% in every frame from the target loss. Four datasets are used for experiments and both qualitative and quantitative results are shown in this paper. Our algorithm succeed the target re-detection in challenging datasets while conventional correlation filter fails.

• Journal of Magnetics
• /
• v.12 no.1
• /
• pp.40-44
• /
• 2007

The Planar Hall effect in a spin valve structure has been applied as a biosensor being capable of detecting $Dynabeads^{(R)}$ M-280. The sensor performance was tested under the application of a DC magnetic field where the output signals were obtained from a nanovoltmeter. The sensor with the pattern size of $50{\times}100{\mu}m^2$ has produced high sensitivity; especially, the real-time profiles by using that sensor revealed significant performance at external applied magnetic field of around 7.0 Oe with the resolution of 0.04 beads per $\mu m^2$. Finally, a successful array including 24 patterns with the single sensor size of $3{\times}3{\mu}m^2$ has shown the uniform and stable signals for single magnetic bead detection. The comparison of this sensor signal with the others has proved feasibility for biosensor application. This, connecting with the advantages of more stable and high signal to noise of PHR sensor's behaviors, can be used to detect the biomolecules and provide a vehicle for detection and study of other molecular interaction.

• Korean Journal of Food Science and Technology
• /
• v.26 no.6
• /
• pp.764-769
• /
• 1994

This study was performed to obtain fast, consistant and reliable estimation system of bacterial counts of raw milk, which effectively related to the quality of sanitaion and the condition of production at the farm. This study compared regression equation and correlation coefficient relationship between standard plate counts and data of Malthus conductance method for the detection time of total, psychrotrophs, coliform bacterial counts in raw milk. Regression equation (RE) between conductance detection time (Y) and total bacterial log counts (X) was Y=18.27651 - 2.07550X, with correlation coefficient -0.95(n=201). In coliform, RE was Y=9.320848 - 1.15598X with correlation coefficient -0.90 (n=207). Psychrotrophs had the RE of Y=29.96008-3.02487 with correlation coeffecient -0.9 (n=201). This conductance method gave results more quickly and was less labor-intensive than traditional standard plate count method.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.614-630
• /
• 2023

Dead trees significantly impact forest production and the ecological environment and pose constraints to the sustainable development of forests. A lightweight YOLOv4 dead tree detection algorithm based on unmanned aerial vehicle images is proposed to address current limitations in dead tree detection that rely mainly on inefficient, unsafe and easy-to-miss manual inspections. An improved logarithmic transformation method was developed in data pre-processing to display tree features in the shadows. For the model structure, the original CSPDarkNet-53 backbone feature extraction network was replaced by MobileNetV3. Some of the standard convolutional blocks in the original extraction network were replaced by depthwise separable convolution blocks. The new ReLU6 activation function replaced the original LeakyReLU activation function to make the network more robust for low-precision computations. The K-means++ clustering method was also integrated to generate anchor boxes that are more suitable for the dataset. The experimental results show that the improved algorithm achieved an accuracy of 97.33%, higher than other methods. The detection speed of the proposed approach is higher than that of YOLOv4, improving the efficiency and accuracy of the detection process.

• Journal of Information Processing Systems
• /
• v.7 no.3
• /
• pp.435-446
• /
• 2011

Microprocessors are becoming increasingly vulnerable to soft errors due to the current trends of semiconductor technology scaling. Traditional redundant multi-threading architectures provide perfect fault tolerance by re-executing all the computations. However, such a full re-execution technique significantly increases the verification workload on the processor resources, resulting in severe performance degradation. This paper presents a pro-active verification management approach to mitigate the verification workload to increase its performance with a minimal effect on overall reliability. An anomaly-speculation-based filter checker is proposed to guide a verification priority before the re-execution process starts. This technique is accomplished by exploiting a value similarity property, which is defined by a frequent occurrence of partially identical values. Based on the biased distribution of similarity distance measure, this paper investigates further application to exploit similar values for soft error tolerance with anomaly speculation. Extensive measurements prove that the majority of instructions produce values, which are different from the previous result value, only in a few bits. Experimental results show that the proposed scheme accelerates the processor to be 180% faster than traditional fully-fault-tolerant processor with a minimal impact on overall soft error rate.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1067-1071
• /
• 2004

In this paper, the performance enhancement algorithm of channel allocation for voice and data transmission in cellular networks is proposed. The voice activity detection has been applied to dynamic channel allocation procedure to detect and separate the silence and speech among conversation periods. Hence a data user can use the silent period of an active voice channel to transmit its information. To control the selecting of channel allocation policies, the information of number of data in transmission waiting queue has been determined in order to accept the performance measurement. In the simulation results, the improvement of the performance shows via the quality of services, which are an average delay in queue, a blocking probability, and an impact of the proposed scheme is presented in the system.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.237-240
• /
• 2003

Gas turbines are extensively used in flight propulsion, electrical power generation, and other industrial applications. During its life span, a turbine blade is taken out periodically for repair and maintenance. This includes re-coating the blade surface and re-drilling the cooling holes/channels. A successful laser re-drilling requires the measurement of a hole within the accuracy of ${\pm}0.15mm$ in position and ${\pm}3^{\circ}$ in orientation. Detection of gas turbine blade/vane cooling hole position and orientation thus becomes a very important step for the vane/blade repair process. The industry is in urgent need of an automated system to fulfill the above task. This paper proposes approaches and algorithms to detect the cooling hole position and orientation by using a vision system mounted on a robot arm. The channel orientation is determined based on the alignment of the vision system with the channel axis. The opening position of the channel is the intersection between the channel axis and the surface around the channel opening. Experimental results have indicated that the concept of cooling hole identification is feasible. It has been shown that the reproducible detection of cooling channel position is with +/- 0.15mm accuracy and cooling channel orientation is with +/$-\;3^{\circ}$ with the current test conditions. Average processing time to search and identify channel position and orientation is less than 1 minute.

• Bulletin of the Korean Chemical Society
• /
• v.34 no.6
• /
• pp.1755-1762
• /
• 2013

A graphene-Zn/Al layered double hydroxide composite film was simultaneously prepared by electrochemical deposition on the surface of a glassy carbon electrode (G-LDH/GCE) from the mixture solution containing GO and nitrate salts of $Zn^{2+}$ and $Al^{3+}$. The modified electrode showed good electrochemical performances toward the simultaneous electrochemical detection of hydroquinone (HQ), catechol (CA) and resorcinol (RE) due to the unique properties of graphene (G) and LDH such as large active surface area, facile electronic transport and high electrocatalytic activity. The redox characteristics of G-LDH/GCE were investigated with cyclic voltammetry and differential pulse voltammetry. The well-separated oxidation peak potentials, corresponding to the oxidation of HQ, CA and RE, were observed at 0.126 V, 0.228 V and 0.620 V respectively. The amperometric response of the modified electrode exhibited that HQ can be detected without interference of CA and RE. Under the optimized conditions, the oxidation peak current of HQ is linear with the concentration of HQ from 6.0 ${\mu}M$ to 325.0 ${\mu}M$ with the detection limit of 0.077 ${\mu}M$ (S/N=3). The modified electrode was successfully applied to the direct determination of HQ in a local tap water, showing reliable recovery data.

• Asian Pacific Journal of Cancer Prevention
• /
• v.16 no.18
• /
• pp.8495-8501
• /
• 2016

Purpose: This study was conducted to assess the agreement and differences between cervical self-sampling with a Kato device (KSSD) and gynecologist sampling for Pap cytology and human papillomavirus DNA (HPV DNA) detection. Materials and Methods: Women underwent self-sampling followed by gynecologist sampling during screening at two primary health clinics. Pap cytology of cervical specimens was evaluated for specimen adequacy, presence of endocervical cells or transformation zone cells and cytological interpretation for cells abnormalities. Cervical specimens were also extracted and tested for HPV DNA detection. Positive HPV smears underwent gene sequencing and HPV genotyping by referring to the online NCBI gene bank. Results were compared between samplings by Kappa agreement and McNemar test. Results: For Pap specimen adequacy, KSSD showed 100% agreement with gynecologist sampling but had only 32.3% agreement for presence of endocervical cells. Both sampling showed 100% agreement with only 1 case detected HSIL favouring CIN2 for cytology result. HPV DNA detection showed 86.2%agreement (K=0.64, 95% CI 0.524-0.756, p=0.001) between samplings. KSSD and gynaecologist sampling identified high risk HPV in 17.3% and 23.9% respectively (p=0.014). Conclusion: The self-sampling using Kato device can serve as a tool in Pap cytology and HPV DNA detection in low resource settings in Malaysia. Self-sampling devices such as KSSD can be used as an alternative technique to gynaecologist sampling for cervical cancer screening among rural populations in Malaysia.