Adaptive Character Segmentation to Improve Text Recognition Accuracy on Mobile Phones

김정식,양형정,김수형,이귀상,김선희,Kim, Jeong Sik,Yang, Hyung Jeong,Kim, Soo Hyung,Lee, Guee Sang,Do, Luu Ngoc,Kim, Sun Hee THE KOREAN INSTITUTE OF SMART MEDIA 2012 스마트미디어저널 Vol.1 No.4

Since mobile phones are used as common communication devices, their applications are increasingly important to human's life. Using smart-phones camera to collect daily life environment's information is one of targets for many applications such as text recognition, object recognition or context awareness. Studies have been conducted to provide important information through the recognition of texts, which are artificially or naturally included in images and movies acquired from mobile phones. In this study, a character segmentation method that improves character-recognition accuracy in images obtained from mobile phone cameras is proposed. The proposed method first classifies texts in a given image to printed letters and handwritten letters since segmentation approaches for them are different. For printed letters, rough segmentation process is conducted, then the segmented regions are integrated, deleted, and re-segmented. Segmentation for the handwritten letters is performed after skews are corrected and the characters are classified by integrating them. The experimental result shows our method achieves a successful performance for both printed and handwritten letters as 95.9% and 84.7%, respectively.

Illumination Invariant Face Tracking on Smart Phones using Skin Locus based CAMSHIFT

Bui, Hoang Nam,Kim, SooHyung,Na, In Seop THE KOREAN INSTITUTE OF SMART MEDIA 2013 스마트미디어저널 Vol.2 No.4

This paper gives a review on three illumination issues of face tracking on smart phones: dark scenes, sudden lighting change and backlit effect. First, we propose a fast and robust face tracking method utilizing continuous adaptive mean shift algorithm (CAMSHIFT) and CbCr skin locus. Initially, the skin locus obtained from training video data. After that, a modified CAMSHIFT version based on the skin locus is accordingly provided. Second, we suggest an enhancement method to increase the chance of detecting faces, an important initialization step for face tracking, under dark illumination. The proposed method works comparably with traditional CAMSHIFT or particle filter, and outperforms these methods when dealing with our public video data with the three illumination issues mentioned above.

Symptoms-Based Power-Efficient Communication Scheme in WBSN

Sasi, Juniven Isin D.,Yang, Hyunho THE KOREAN INSTITUTE OF SMART MEDIA 2014 스마트미디어저널 Vol.3 No.1

It is practical nowadays to automate data recording in order to prevent loss and tampering of records. There are existing technologies that satisfy this needs and one of them is wireless sensor networks (WSN). Wireless body sensor networks (WBSN) are wireless networks and information-processing systems which are deployed to monitor medical condition of patients. In terms of performance, WBSNs are restricted by energy, and communication between nodes. In this paper, we focused in improving the performance of communication to achieve less energy consumption and to save power. The main idea of this paper is to prioritize nodes that exhibit a sudden change of vital signs that could put the patient at risk. Cluster head is the main focus of this study in order to be effective; its main role is to check the sent data of the patient that exceeds threshold then transfer to the sink node. The proposed scheme implemented added a time-based protocol to sleep/wakeup mechanism for the sensor nodes. We seek to achieve a low energy consumption and significant throughput in this study.

Noise Removal for Level Set based Flower Segmentation

박상철,오강한,나인섭,김수형,양형정,이귀상,Park, Sang Cheol,Oh, Kang Han,Na, In Seop,Kim, Soo Hyung,Yang, Hyung Jeong,Lee, Guee Sang THE KOREAN INSTITUTE OF SMART MEDIA 2012 스마트미디어저널 Vol.1 No.2

본 연구에서는 노이즈를 제거하고 자연 영상에서 자동으로 꽃을 분할하는 후처리방법을 제시한다. 레벨 셋 알고리즘을 이용한 자연영상 꽃 분할에서는 레벨 셋이 에지 정보에만 의존하기 때문에 기대하지 않았던 분리된 노이즈들이 발생한다. 실험 결과는 제안 방법이 꽃 영역과 배경 영역의 많은 노이즈를 성공적으로 제거하였음을 보여준다. In this paper, post-processing step is presented to remove noises and develop a fully automated scheme to segment flowers in natural scene images. The scheme to segment flowers using a level set algorithm in the natural scene images produced unexpected and isolated noises because the level set relies only on the color and edge information. The experimental results shows that the proposed method successfully removes noises in the foreground and background.

Feature Selection and Performance Analysis using Quantum-inspired Genetic Algorithm

허기수,정현태,박아론,백성준,Heo, G.S.,Jeong, H.T.,Park, A.,Baek, S.J. THE KOREAN INSTITUTE OF SMART MEDIA 2012 스마트미디어저널 Vol.1 No.1

특징 선택은 패턴 인식의 성능을 향상시키기 위해 부분집합을 구성하는 중요한 문제다. 특징 선택에는 순차 탐색 알고리즘으로부터 확률 기반의 유전 알고리즘까지 다양한 접근 방법이 적용 되었다. 본 연구에서는 특징 선택을 위해 양자 비트, 상태의 중첩 등 양자 컴퓨터 개념을 기반으로 하는 양자 기반 유전 알고리즘(QGA: Quantum-inspired Genetic Algorithm)을 적용하였다. QGA 성능은 전통적인 유전 알고리즘(CGA: Conventional Genetic Algorithm)을 적용한 특징 선택 방법과 분류율 및 평균 특징 개수의 비교를 통해 이루어졌으며, UCI 데이터를 이용한 실험 결과 QGA를 적용한 특징 선택 방법이 CGA를 적용한 경우에 비해 전반적으로 좋은 성능을 보임을 확인 할 수 있었다. Feature selection is the important technique of selecting a subset of relevant features for building robust pattern recognition systems. Various methods have been studied for feature selection from sequential search algorithms to stochastic algorithms. In this work, we adopted a Quantum-inspired Genetic Algorithm (QGA) which is based on the concept and principles of quantum computing such as Q-bits and superposition of state for feature selection. The performance of QGA is compared to that of the Conventional Genetic Algorithm (CGA) with respect to the classification rates and the number of selected features. The experimental result using UCI data sets shows that QGA is superior to CGA.

Implementation of Wireless PGN Analyzer for ISOBUS network

Tumenjargal, Enkhbaatar,Badarch, Luubaatar,Lee, Kangsan,Ham, Woonchul,Doopalam, Enkhzul,Togooch, Amartuvshin THE KOREAN INSTITUTE OF SMART MEDIA 2015 스마트미디어저널 Vol.4 No.2

Communication between ECUs (Electronic Control Units) in agricultural machineries tends to use IS011783 widely, that is PGN (Parameter Group Number) based communication protocol lays on CAN protocol by altering its identifier part. Messages in line are transferred and received between ECUs according to ISO11783 standard. This paper discusses about design of wireless monitoring system. We used an ARM Cortex-M3 microcontroller embedded development board and marvel8686 wireless module. The wireless ISOBUS monitoring system, attached to communication line, reads messages, interpret them, and display them on the screen in easily comprehendible form. It can be used to generate messages and monitor the traffic on physical bus systems. The monitoring system connected to ECUs, monitor and simulate real traffic of communication and functionality of the ECUs. In order to support our work, we have implemented the monitoring tool. The development consists of two parts: GUI of the application and firmware level programming. Hence the monitoring system is attached to the communication line and equipped by Wi-Fi module; farmer/dispatcher in a farm monitors all messages in communication line on personal computer and smart device.

Animal Tracking in Infrared Video based on Adaptive GMOF and Kalman Filter

Pham, Van Khien,Lee, Guee Sang THE KOREAN INSTITUTE OF SMART MEDIA 2016 스마트미디어저널 Vol.5 No.1

The major problems of recent object tracking methods are related to the inefficient detection of moving objects due to occlusions, noisy background and inconsistent body motion. This paper presents a robust method for the detection and tracking of a moving in infrared animal videos. The tracking system is based on adaptive optical flow generation, Gaussian mixture and Kalman filtering. The adaptive Gaussian model of optical flow (GMOF) is used to extract foreground and noises are removed based on the object motion. Kalman filter enables the prediction of the object position in the presence of partial occlusions, and changes the size of the animal detected automatically along the image sequence. The presented method is evaluated in various environments of unstable background because of winds, and illuminations changes. The results show that our approach is more robust to background noises and performs better than previous methods.

Analyze weeds classification with visual explanation based on Convolutional Neural Networks

Vo, Hoang-Trong,Yu, Gwang-Hyun,Nguyen, Huy-Toan,Lee, Ju-Hwan,Dang, Thanh-Vu,Kim, Jin-Young THE KOREAN INSTITUTE OF SMART MEDIA 2019 스마트미디어저널 Vol.8 No.3

To understand how a Convolutional Neural Network (CNN) model captures the features of a pattern to determine which class it belongs to, in this paper, we use Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize and analyze how well a CNN model behave on the CNU weeds dataset. We apply this technique to Resnet model and figure out which features this model captures to determine a specific class, what makes the model get a correct/wrong classification, and how those wrong label images can cause a negative effect to a CNN model during the training process. In the experiment, Grad-CAM highlights the important regions of weeds, depending on the patterns learned by Resnet, such as the lobe and limb on 미국가막사리, or the entire leaf surface on 단풍잎돼지풀. Besides, Grad-CAM points out a CNN model can localize the object even though it is trained only for the classification problem.

Tracking by Detection of Multiple Faces using SSD and CNN Features

Tai, Do Nhu,Kim, Soo-Hyung,Lee, Guee-Sang,Yang, Hyung-Jeong,Na, In-Seop,Oh, A-Ran THE KOREAN INSTITUTE OF SMART MEDIA 2018 스마트미디어저널 Vol.7 No.4

Multi-tracking of general objects and specific faces is an important topic in the field of computer vision applicable to many branches of industry such as biometrics, security, etc. The rapid development of deep neural networks has resulted in a dramatic improvement in face recognition and object detection problems, which helps improve the multiple-face tracking techniques exploiting the tracking-by-detection method. Our proposed method uses face detection trained with a head dataset to resolve the face deformation problem in the tracking process. Further, we use robust face features extracted from the deep face recognition network to match the tracklets with tracking faces using Hungarian matching method. We achieved promising results regarding the usage of deep face features and head detection in a face tracking benchmark.

A Nearly Optimal One-to-Many Routing Algorithm on k-ary n-cube Networks

Choi, Dongmin,Chung, Ilyong THE KOREAN INSTITUTE OF SMART MEDIA 2018 스마트미디어저널 Vol.7 No.2

The k-ary n-cube $Q^k_n$ is widely used in the design and implementation of parallel and distributed processing architectures. It consists of $k^n$ identical nodes, each node having degree 2n is connected through bidirectional, point-to-point communication channels to different neighbors. On $Q^k_n$ we would like to transmit packets from a source node to 2n destination nodes simultaneously along paths on this network, the $i^{th}$ packet will be transmitted along the $i^{th}$ path, where $0{\leq}i{\leq}2n-1$. In order for all packets to arrive at a destination node quickly and securely, we present an $O(n^3)$ routing algorithm on $Q^k_n$ for generating a set of one-to-many node-disjoint and nearly shortest paths, where each path is either shortest or nearly shortest and the total length of these paths is nearly minimum since the path is mainly determined by employing the Hungarian method.