• 융합보안논문지
• /
• 제12권6호
• /
• pp.31-37
• /
• 2012

오늘날 자동차 기술의 핵심은 IT 기반 융합 시스템기술로 변화하고 있다. 다양한 IT 기술을 접목하여 운전 중 다양한 상황에 대응하고 또한 운전자의 편의성을 지원하는 기술적 추세를 보이고 있다. 본 논문에서는 운전자의 안전성과 편의성을 증대하기 위해 영상 정보를 기반으로 도로 정보를 검출해 운전자에게 알려주고, 버튼을 직접 손으로 눌러야 하는 물리적 인터페이스를 대체할 비접촉식 인터페이스 기술을 융합한 Augmented Driving System (ADS) 기술을 제안한다. 본 기술은 카메라로부터 입력 받은 영상 정보를 제안된 알고리즘을 통해 앞차와의 거리, 차선, 교통 표지판을 검출하고 차량 내부를 주시하는 카메라와 운전자의 음성을 인식할 마이크를 기반으로 기본 음성인식과 동작인식이 융합된 인터페이스 기술을 제공한다. 이러한 요소 기술들은 운전자가 인지하지 못하더라도 운전자에게 현재의 주행상황을 인지하여 자동으로 알려줌으로써 교통사고 확률을 크게 낮출 수 있을 것이며, 또한 다양한 운전 중 기능 조작을 편리하게 지원함으로써 운전자의 전방 주시에 도움을 줄 수 있다. 본 논문에서 개발된 기술을 통해 테스트를 실시해 본 결과 표지판인식, 차선검출, 앞차와의 거리 검출 등의 인식률이 약 90% 이상이 되었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권7호
• /
• pp.2547-2567
• /
• 2021

It has been widely acknowledged that occlusion impairments adversely distress many face recognition algorithms' performance. Therefore, it is crucial to solving the problem of face image occlusion in face recognition. To solve the image occlusion problem in face recognition, this paper aims to automatically de-occlude the human face majority or discriminative regions to improve face recognition performance. To achieve this, we decompose the generative process into two key stages and employ a separate generative adversarial network (GAN)-based network in both stages. The first stage generates an initial coarse face image without an occlusion mask. The second stage refines the result from the first stage by forcing it closer to real face images or ground truth. To increase the performance and minimize the artifacts in the generated result, a new refine loss (e.g., reconstruction loss, perceptual loss, and adversarial loss) is used to determine all differences between the generated de-occluded face image and ground truth. Furthermore, we build occluded face images and corresponding occlusion-free face images dataset. We trained our model on this new dataset and later tested it on real-world face images. The experiment results (qualitative and quantitative) and the comparative study confirm the robustness and effectiveness of the proposed work in removing challenging occlusion masks with various structures, sizes, shapes, types, and positions.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2016년도 춘계학술대회
• /
• pp.359-362
• /
• 2016

얼굴은 홍채, 지문 등과 같은 사람마다 가진 특수한 정보이다. 얼굴 인식에 대한 연구들은 과거부터 현재까지 지속적으로 진행되고 있으며, 이러한 연구들을 통해 여러 가지의 얼굴 인식 방법들이 나타났다. 이 중에는 스테레오로 구성된 얼굴 데이터를 이용하여 얼굴 인식을 진행하는 알고리즘들이 있다. 본 논문에서는 기계학습의 방법인 Convolutional Neural Network를 이용하여 스테레오로 구성된 얼굴 이미지를 하나의 신경망으로 학습을 진행하였다. 또한 스테레오로 구성된 얼굴 이미지는 카메라 2대를 이용하여 취득하였다. 이 방법은 얼굴 인식에서 보편적으로 많이 사용되는 알고리즘인 PCA를 이용한 스테레오 얼굴 인식의 결과보다 더욱 좋은 성능을 보였다.

• International Journal of Advanced Culture Technology
• /
• 제7권2호
• /
• pp.209-217
• /
• 2019

Due to the development of artificial intelligence and image recognition technology that play important roles in the field of 4th industry, office automation systems and unmanned automation systems are rapidly spreading in human society. The proposed algorithm first finds the variances of the differences between the tile values constituting the learning characters and the experimental character and then recognizes the experimental character according to the distribution of the three learning characters with the smallest variances. In more detail, for 100 learning data characters and 10 experimental data characters, each character is defined as the number of black pixels belonging to 15 tile areas. For each character constituting the experimental data, the variance of the differences of the tile values of 100 learning data characters is obtained and then arranged in the ascending order. After that, three learning data characters with the minimum variance values are selected, and the final recognition result for the given experimental character is selected according to the distribution of these character types. Moreover, we compare the recognition result with the result made by a neural network of basic structure. It is confirmed that satisfactory recognition results are obtained through the processes that subdivide the learning characters and experiment characters into tile sizes and then select the recognition result using variances.

• Journal of Platform Technology
• /
• 제12권1호
• /
• pp.3-11
• /
• 2024

Deaf people who use sign language as their first language sometimes have difficulty communicating because they do not know spoken Korean. Deaf people are also members of society, so we must support to create a society where everyone can live together. In this paper, we present a method to increase the recognition rate of Korean sign language using a CNN model. When the original image was used as input to the CNN model, the accuracy was 0.96, and when the image corresponding to the skin area in the YCbCr color space was used as input, the accuracy was 0.72. It was confirmed that inserting the original image itself would lead to better results. In other studies, the accuracy of the combined Conv1d and LSTM model was 0.92, and the accuracy of the AlexNet model was 0.92. The CNN model proposed in this paper is 0.96 and is proven to be helpful in recognizing Korean sign language.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권1호
• /
• pp.378-390
• /
• 2015

Image projectors can turn any surface into a display. Integrating a surface projection with a user interface transforms it into an interactive display with many possible applications. Hand gesture interfaces are often used with projector-camera systems. Hand detection through color image processing is affected by the surrounding environment. The lack of illumination and color details greatly influences the detection process and drops the recognition success rate. In addition, there can be interference from the projection system itself due to image projection. In order to overcome these problems, a gesture interface based on depth images is proposed for projected surfaces. In this paper, a depth camera is used for hand recognition and for effectively extracting the area of the hand from the scene. A hand detection and finger tracking method based on depth images is proposed. Based on the proposed method, a touch interface for the projected surface is implemented and evaluated.

• 한국항해항만학회:학술대회논문집
• /
• 한국항해항만학회 2023년도 춘계학술대회
• /
• pp.40-41
• /
• 2023

본 논문에서는 VTS 레이더 이미지 기반 객체의 탐지, 인식, 추적 알고리즘의 설계에 대해 소개한다. 레이더 이미지 기반 객체 탐지는 인공지능 기술을 이용하여 객체 유무 여부를 확인하고, 탐지의 경우 인공지능 기술을 이용하여 선종을 구분하게 된다. 추적은 탐지된 객체에 대해 시간에 따른 연속적 추적을 실시하며 이동경로의 혼선을 방지하는 기술이 포함되어 있다. 특히 육상레이더의 경우 지형지물에 따라 탐지가 불필요한 영역이 있어, 레이더 이미지에서 관심영역(ROI)을 설정하여 영역 내 선박을 탐지하고 인식하는 기능이 포함되어 있다. 또한, 추출한 좌표정보를 통해 속도와 방향 등을 계산하여 다양한 응용 해석이 가능하도록 설계하였다.

• Journal of Information Processing Systems
• /
• 제19권4호
• /
• pp.427-438
• /
• 2023

Plenty of works have indicated that single image super-resolution (SISR) models relying on synthetic datasets are difficult to be applied to real scene text image super-resolution (STISR) for its more complex degradation. The up-to-date dataset for realistic STISR is called TextZoom, while the current methods trained on this dataset have not considered the effect of multi-scale features of text images. In this paper, a multi-scale and attention fusion model for realistic STISR is proposed. The multi-scale learning mechanism is introduced to acquire sophisticated feature representations of text images; The spatial and channel attentions are introduced to capture the local information and inter-channel interaction information of text images; At last, this paper designs a multi-scale residual attention module by skillfully fusing multi-scale learning and attention mechanisms. The experiments on TextZoom demonstrate that the model proposed increases scene text recognition's (ASTER) average recognition accuracy by 1.2% compared to text super-resolution network.

• 산업기술연구
• /
• 제17권
• /
• pp.331-338
• /
• 1997

In this paper, we propose a new neural pattern recognition from wavelet transform. We first analysis in BFT(Binary Field Transform) in character image. The proposed neural network and wavelet transform is able to improve learning time and scaling. The ability and effectiveness of identifying image using the proposed wavelet transform will be demonstrated by computer simulation.

• Journal of information and communication convergence engineering
• /
• 제12권4호
• /
• pp.263-270
• /
• 2014

Corner detection and feature extraction are essential aspects of computer vision problems such as object recognition and tracking. Feature detectors such as Scale Invariant Feature Transform (SIFT) yields high quality features but computationally intensive for use in real-time applications. The Features from Accelerated Segment Test (FAST) detector provides faster feature computation by extracting only corner information in recognising an object. In this paper we have analyzed the efficient object detection algorithms with respect to efficiency, quality and robustness by comparing characteristics of image detectors for corner detector and feature extractors. The simulated result shows that compared to conventional SIFT algorithm, the object recognition system based on the FAST corner detector yields increased speed and low performance degradation. The average time to find keypoints in SIFT method is about 0.116 seconds for extracting 2169 keypoints. Similarly the average time to find corner points was 0.651 seconds for detecting 1714 keypoints in FAST methods at threshold 30. Thus the FAST method detects corner points faster with better quality images for object recognition.