• Journal of Web Engineering
• /
• v.21 no.4
• /
• pp.1283-1294
• /
• 2022

Traffic volume is gradually increasing due to the development of technology and the concentration of people in cities. As the results, traffic congestion and traffic accidents are becoming social problems. Detecting and tracking a vehicle based on computer vision is a great helpful in providing important information such as identifying road traffic conditions and crime situations. However, vehicle detection and tracking using a camera is affected by environmental factors in which the camera is installed. In this paper, we thus propose a deep learning based on vehicle classification and tracking scheme to classify and track vehicles in a complex and diverse environment. Using YOLO model as deep learning model, it is possible to quickly and accurately perform robust vehicle tracking in various environments, compared to the traditional method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.12
• /
• pp.6009-6027
• /
• 2019

Support vector machine (SVM) classifiers have been widely used for object detection. These methods usually locate the object by finding the region with maximal score in an image. With bag-of-features representation, the SVM score of an image region can be written as the sum of its inside feature-weights. As a result, the searching process can be executed efficiently by using strategies such as branch-and-bound. However, the feature-weight derived by optimizing region classification cannot really reveal the category knowledge of a feature-point, which could cause bad localization. In this paper, we represent a region in an image by a collection of local feature-points and determine the object by the region with the maximum posterior probability of belonging to the object class. Based on the Bayes' theorem and Naive-Bayes assumptions, the posterior probability is reformulated as the sum of feature-scores. The feature-score is manifested in the form of the logarithm of a probability ratio. Instead of estimating the numerator and denominator probabilities separately, we readily employ the density ratio estimation techniques directly, and overcome the above limitation. Experiments on a car dataset and PASCAL VOC 2007 dataset validated the effectiveness of our method compared to the baselines. In addition, the performance can be further improved by taking advantage of the recently developed deep convolutional neural network features.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.1311-1314
• /
• 2022

DNN을 사용하여 객체 인식 과정에서 객체를 잘 분류하기 위해서는 시각적 설명성이 요구된다. 시각적 설명성은 object class에 대한 예측을 pixel-wise attribution으로 표현해 예측 근거를 해석하기 위해 제안되었다, Scale-invariant한 특징을 제공하도록 설계된 pyramidal features 기반 backbone 구조는 object detection 및 classification 등에서 널리 쓰이고 있으며, 이러한 특징을 갖는 feature pyramid를 trainable attention mechanism에 적용하고자 할 때 계산량 및 메모리의 복잡도가 증가하는 문제가 있다. 본 논문에서는 일반적인 FPN에서 객체 인식 성능과 설명성을 높이기 위한 피라미드-주의집중 계층네트워크 (FPN-Attention Layered Network) 방식을 제안하고, 실험적으로 그 특성을 평가하고자 한다. 기존의 FPN만을 사용하였을 때 객체 인식 과정에서 설명성을 향상시키는 방식이 객체 인식에 미치는 정도를 정량적으로 평가하였다. 제안된 모델의 적용을 통해 낮은 computing 오버헤드 수준에서 multi-level feature를 고려한 시각적 설명성을 개선시켜, 결괴적으로 객체 인식 성능을 향상 시킬 수 있음을 실험적으로 확인할 수 있었다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2016.07a
• /
• pp.51-54
• /
• 2016

In computer vision especially in image processing, it has become popular to apply deep convolutional networks for supervised learning. Convolutional networks have shown a state of the art results in classification, object recognition, detection as well as semantic segmentation. However, supervised learning has two major disadvantages. One is it requires huge amount of labeled data to get high accuracy, the second one is to train so much data takes quite a bit long time. On the other hand, unsupervised learning can handle these problems more cheaper way. In this paper we show efficient way to learn features for classification in an unsupervised way. The network trained layer-wise, used backpropagation and our network learns features from unlabeled data. Our approach shows better results on Caltech-256 and STL-10 dataset.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.10
• /
• pp.137-148
• /
• 2024

The European honeybee (Apis mellifera) is an important pollinator threatened by colony collapse disorder (CCD), primarily due to infestation by the Varroa mite (Varroa destructor). Traditional detection methods are invasive and time-consuming, often causing additional stress to colonies. We propose a two-stage framework using the You Only Look Once version 8 (YOLOv8) model for non-destructive and rapid detection of Varroa mite infestation. The framework uses comb light images from inside the hives. In the first stage, a YOLOv8-n model detects bees and extracts individual bee images. In the second stage, a YOLOv8-cls model classifies the infestation status of each bee. Our object detection model achieved a mAP@0.5 of 0.701, and the classification model achieved an average accuracy of 91%. These results demonstrate the effectiveness of the framework as a non-destructive method for Varroa mite detection. Based on this research, we expect to provide beekeepers with an efficient tool for early detection and management of Varroa mite infestations, potentially reducing the incidence of CCD and supporting the sustainability of apiculture.

• Proceedings of the KSRS Conference
• /
• 2006.03a
• /
• pp.307-310
• /
• 2006

고해상도(VHR, Very High Resolution) 영상은 활용에 따라 도심의 다양한 정보를 얻을 수 있는 잠재적 가치가 매우 큰 자료이다. 그러나 이러한 고해상도 영상자료는 매우 높은 공간해상력으로 인해 같은 용도의 객체 혹은 같은 객체(예, 건물)라 할지라도 다양한 분광 특성 및 형태로 표현된다. 그러므로 이러한 고해상도영상을 이용하여 효과적으로 주제도를 생성하기 위해서는 현재까지 영상분류 분야에서 주로 활용되고 있는 화소(pixel)단위 기반의 분석방법으로는 한계가 존재한다. 본 연구에서는 이러한 문제점을 보완하기 위한 방법으로 활발한 연구가 진행되고 있는 세그멘트(segment) 혹은 객체(object) 기반 분류기법을 고해상도 영상 및 라이다 자료에 적용하여 도심지역의 건물들을 추출해 보았으며, 그 활용 가능성에 대하여 판단해 보았다. 이러한 세그멘트 기법은 분류하고자 하는 객체들을 하나의 동일한 특성을 가지는 집단으로 모으는 방법을 말하는데, 이를 위해 본 연구에서는 multi-resolution image segmentation기법을 제공해주는 eCognition이라는 소프트웨어를 이용하였다.

• Journal of Cadastre & Land InformatiX
• /
• v.47 no.2
• /
• pp.79-90
• /
• 2017

Urban environment represent one of the most dynamic regions on earth. As in other countries, forests, green areas, agricultural lands are rapidly changing into residential or industrial areas in South Korea. Monitoring such rapid changes in land use requires rapid data acquisition, and satellite imagery can be an effective method to this demand. In general, SAR(Synthetic Aperture Radar) satellites acquire images with an active system, so the brightness of the image is determined by the surface roughness. Therefore, the water areas appears dark due to low reflection intensity, In the residential area where the artificial structures are distributed, the brightness value is higher than other areas due to the strong reflection intensity. If we use these characteristics of SAR images, settlement areas can be extracted efficiently. In this study, extraction of settlement areas was performed using TerraSAR-X of German high-resolution X-band SAR satellite and KOMPSAT-5 of South Korea, and object-oriented image classification method using the image segmentation technique is applied for extraction. In addition, to improve the accuracy of image segmentation, the speckle divergence was first calculated to adjust the reflection intensity of settlement areas. In order to evaluate the accuracy of the two satellite images, settlement areas are classified by applying a pixel-based K-means image classification method. As a result, in the case of TerraSAR-X, the accuracy of the object-oriented image classification technique was 88.5%, that of the pixel-based image classification was 75.9%, and that of KOMPSAT-5 was 87.3% and 74.4%, respectively.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.653-657
• /
• 2005

AGV is very useful equipment to transfer containers in automated container terminal. AGV must have Obstacle Detection System (ODS) for port automation. ODS needs the function to classify some specified object from background in acquired data. And it must be able to track classified moving objects. Finally, ODS could determine its next action for safe driving whether it should do emergency stop or speed down, or it should change its deriving lane. For these functions, ODS can have many different kinds of algorithm. In this paper, we present one of AGV to be used in automated container terminal.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.46 no.4
• /
• pp.101-106
• /
• 2023

With the recent surge in YouTube usage, there has been a proliferation of user-generated videos where individuals evaluate cosmetics. Consequently, many companies are increasingly utilizing evaluation videos for their product marketing and market research. However, a notable drawback is the manual classification of these product review videos incurring significant costs and time. Therefore, this paper proposes a deep learning-based cosmetics search algorithm to automate this task. The algorithm consists of two networks: One for detecting candidates in images using shape features such as circles, rectangles, etc and Another for filtering and categorizing these candidates. The reason for choosing a Two-Stage architecture over One-Stage is that, in videos containing background scenes, it is more robust to first detect cosmetic candidates before classifying them as specific objects. Although Two-Stage structures are generally known to outperform One-Stage structures in terms of model architecture, this study opts for Two-Stage to address issues related to the acquisition of training and validation data that arise when using One-Stage. Acquiring data for the algorithm that detects cosmetic candidates based on shape and the algorithm that classifies candidates into specific objects is cost-effective, ensuring the overall robustness of the algorithm.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.1
• /
• pp.85-92
• /
• 2020

Deep learning shows outstanding performance in image and video analysis, such as object classification, object detection and semantic segmentation. In this paper, it is analyzed that the performances of deep learning models can be affected by characteristics of train dataset. It is proposed as a method for selecting activation function and optimization algorithm of deep learning to classify facial expression. Classification performances are compared and analyzed by applying various algorithms of each component of deep learning model for CK+, MMI, and KDEF datasets. As results of simulation, it is shown that genetic algorithm can be an effective solution for optimizing components of deep learning model.