• 한국멀티미디어학회논문지
• /
• 제24권10호
• /
• pp.1358-1368
• /
• 2021

A research on agricultural automation is a main issues to overcome the shortage of labor in Korea. A sweet persimmon farmers need much time and labors for classifying profitable sweet persimmon and ill profitable products. In this paper, we propose a mixed two-step synthetic neural network model for efficiently classifying sweet persimmon images. In this model, we suggested a surface direction classification model and a quality screening model which constructed from image data sets. Also we studied Class Activation Mapping(CAM) for visualization to easily inspect the quality of the classified products. The proposed mixed two-step model showed high performance compared to the simple binary classification model and the multi-class classification model, and it was possible to easily identify the weak parts of the classification in a dataset.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2003년도 Proceedings of ACRS 2003 ISRS
• /
• pp.865-865
• /
• 2003

• 한국지능시스템학회논문지
• /
• 제23권3호
• /
• pp.226-231
• /
• 2013

의약품을 자동 포장하는 시스템에서는 캐니스터(Canister)에 해당 약을 정확히 보충할 수 있는 해당 약통과 캐니스터와의 일치 여부를 판단하는 정합 알고리즘이 필수적이다. 본 논문에서는 약화사고 방지를 위해 많은 종류의 약통을 분류하기 위한 분류 성능뿐만 아니라 실시간으로 처리할 수 있는 상 하 계층으로 구성된 계층적 신경회로망을 제안한다. 먼저 약통 정보를 나타내는 라벨 영상으로부터 다수의 저 차원 특징 벡터를 추출한다. 추출된 특징 벡터를 사용하여 하위계층의 다층 퍼셉트론(MLP, Multi-layer Perceptron) 신경회로망을 학습한다. 다음으로 학습된 MLP의 중간층 출력을 입력으로 사용하여 상위계층의 MLP를 학습한다. 100개의 약통에 대해 좌우 30도까지 회전한 영상에 대해 제안한 계층적 신경회로망의 분류 성능 시험과 실시간 연산처리 성능의 우수함을 보였다.

• 대한조선학회논문집
• /
• 제57권3호
• /
• pp.140-151
• /
• 2020

It is important to identify the location of ship hull blocks with exact block identification number when scheduling the shipbuilding process. The wrong information on the location and identification number of some hull block can cause low productivity by spending time to find where the exact hull block is. In order to solve this problem, it is necessary to equip the system to track the location of the blocks and to identify the identification numbers of the blocks automatically. There were a lot of researches of location tracking system for the hull blocks on the stockyard. However there has been no research to identify the hull blocks on the stockyard. This study compares the performance of 5 Convolutional Neural Network (CNN) models with multi-view image set on the classification of the hull blocks to identify the blocks on the stockyard. The CNN models are open algorithms of ImageNet Large-Scale Visual Recognition Competition (ILSVRC). Four scaled hull block models are used to acquire the images of ship hull blocks. Learning and transfer learning of the CNN models with original training data and augmented data of the original training data were done. 20 tests and predictions in consideration of five CNN models and four cases of training conditions are performed. In order to compare the classification performance of the CNN models, accuracy and average F1-Score from confusion matrix are adopted as the performance measures. As a result of the comparison, Resnet-152v2 model shows the highest accuracy and average F1-Score with full block prediction image set and with cropped block prediction image set.

• 대한원격탐사학회지
• /
• 제22권1호
• /
• pp.25-40
• /
• 2006

Synthetic aperture radar (SAR) from satellites provides the opportunity to regularly incorporate microwave information into forest classification. Radar backscatter can improve classification accuracy, and SAR interferometry could provide improved thematic information through the use of coherence. This research examined the potential of using multi-temporal JERS-l SAR (L band) backscatter information and interferometry in distinguishing forest classes of mountainous areas in the Northeastern U.S. for future forest mapping and monitoring. Raw image data from a pair of images were processed to produce coherence and backscatter data. To improve the geometric characteristics of both the coherence and the backscatter images, this study used the interferometric techniques. It was necessary to radiometrically correct radar backscatter to account for the effect of topography. This study developed a simplified method of radiometric correction for SAR imagery over the hilly terrain, and compared the forest-type discriminatory powers of the radar backscatter, the multi-temporal backscatter, the coherence, and the backscatter combined with the coherence. Statistical analysis showed that the method of radiometric correction has a substantial potential in separating forest types, and the coherence produced from an interferometric pair of images also showed a potential for distinguishing forest classes even though heavily forested conditions and long time separation of the images had limitations in the ability to get a high quality coherence. The method of combining the backscatter images from two different dates and the coherence in a multivariate approach in identifying forest types showed some potential. However, multi-temporal analysis of the backscatter was inconclusive because leaves were not the primary scatterers of a forest canopy at the L-band wavelengths. Further research in forest classification is suggested using diverse band width SAR imagery and fusing with other imagery source.

• 전기전자학회논문지
• /
• 제26권3호
• /
• pp.364-372
• /
• 2022

본 논문에서는 FMCW(frequency modulated continuous wave) 레이다 센서를 활용한 사람과 사물을 분류하는 시스템 설계 및 구현 결과를 제시한다. 해당 시스템은 다중 객체 탐지를 위한 레이다 센서 신호처리 과정과 객체를 사람 및 사물로 분류하는 딥러닝 과정을 수행한다. 딥러닝의 경우 높은 연산량과 많은 양의 메모리를 요구하기 때문에 경량화가 필수적이다. 따라서 CNN (convolution neural network) 연산을 이진화하여 동작하는 BNN (binary neural network) 구조를 적용하였으며, 실시간 동작을 위해 하드웨어 가속기를 설계하고 FPGA 보드 상에서 구현 및 검증하였다. 성능 평가 및 검증 결과 90.5%의 다중 객체 구분 정확도, CNN 대비 96.87% 감소된 메모리 구현이 가능하며, 총 수행 시간은 5ms로 실시간 동작이 가능함을 확인하였다.

• 한국컴퓨터정보학회논문지
• /
• 제23권9호
• /
• pp.7-13
• /
• 2018

Recent advancement in data gathering technique improves the capability of information collecting, thus allowing the learning process between gathered data patterns and application sub-tasks. A pattern can be associated with multiple labels, demanding multi-label learning capability, resulting in significant attention to multi-label feature selection since it can improve multi-label learning accuracy. However, existing evolutionary multi-label feature selection methods suffer from ineffective search process. In this study, we propose a evolutionary search process for the task of multi-label feature selection problem. The proposed method creates large set of offspring or new feature subsets and then retains the most promising feature subset. Experimental results demonstrate that the proposed method can identify feature subsets giving good multi-label classification accuracy much faster than conventional methods.

• 한국HCI학회:학술대회논문집
• /
• 한국HCI학회 2009년도 학술대회
• /
• pp.21-26
• /
• 2009

유비쿼터스 컴퓨팅 환경의 급속한 발전은 Multi-Sensor를 이용하여 자동으로 사용자의 행동인식을 가능한 환경을 만들어주었다. 따라서 이 논문에서는 사용자가 일상생활을 하는데 있어서 기본적으로 필요한 행동인 ADL(activities of daily living)의 수행능력을 분석하고 진단할 수 있는 Multi-Sensor기반의 ADL 자동 진단 시스템을 구축하였다. 두 개의 가속도 센서를 허벅지와 손목에 부착하여 사용자의 행동 정보를 수집하고 이를 Decision-Tree를 통하여 분석하여 사용자의 행동 정보를 수집하였다. 또한 Zigbee 센서를 이용하여 개별 물체의 Object ID를 이용하여 사용자의 위치정보와 주변의 물체의 정보를 수집하여 사용자의 상태 정보를 수집하였다. 이렇게 수집된 행동 정보와 상태 정보들을 통하여 일상생활에 필요한 약 20여 가지의 행동을 인식하였고 평균적으로 96%이상의 정확도를 나타내었으며 이를 통하여 ADL 지수를 자동으로 측정하였다.

• 제어로봇시스템학회논문지
• /
• 제19권4호
• /
• pp.364-371
• /
• 2013

Emotion Recognition is one of the important part to develop in human-human and human computer interaction. In this paper, we have focused on the performance of multi-class SVM (Support Vector Machine) with Gaussian RFB (Radial Basis function) kernel, which has been used to solve the problem of emotion recognition from physiological signals and to improve the accuracy of emotion recognition. The experimental paradigm for data acquisition, visual-stimuli of IAPS (International Affective Picture System) are used to induce emotional states, such as fear, disgust, joy, and neutral for each subject. The raw signals of acquisited data are splitted in the trial from each session to pre-process the data. The mean value and standard deviation are employed to extract the data for feature extraction and preparing in the next step of classification. The experimental results are proving that the proposed approach of multi-class SVM with Gaussian RBF kernel with OVO (One-Versus-One) method provided the successful performance, accuracies of classification, which has been performed over these four emotions.

• 한국정밀공학회지
• /
• 제26권11호
• /
• pp.124-130
• /
• 2009

In the multi-stage shape drawing process, the most important aspect for the economy is the correct design of the various drawing stage. For most of the products commonly available round or square materials can be used as initial material. However, special products should be pre-rolled. This study proposes a process design method of multi-stage shape drawing process for producing cross roller guide. Firstly, a standard classification of shape drawing process is suggested based on the requirement of pre-rolling process. And a design method is proposed to design the intermediate die shape. The process design method is applied to design the multi-stage shape drawing process for producing cross roller guide. Finally, the effectiveness of the proposed design method is verified by FE-analysis and shape drawing experiment.