• Journal of Multimedia Information System
• /
• 제9권1호
• /
• pp.1-10
• /
• 2022

China is a big country in animal fur industry. The total production and consumption of fur are increasing year by year. However, the recognition of fur in the fur production process still mainly relies on the visual identification of skilled workers, and the stability and consistency of products cannot be guaranteed. In response to this problem, this paper proposes a feature fusion-based animal fur recognition network on the basis of typical convolutional neural network structure, relying on rapidly developing deep learning techniques. This network superimposes texture feature - the most prominent feature of fur image - into the channel dimension of input image. The output feature map of the first layer convolution is inverted to obtain the inverted feature map and concat it into the original output feature map, then Leaky ReLU is used for activation, which makes full use of the texture information of fur image and the inverted feature information. Experimental results show that the algorithm improves the recognition accuracy by 9.08% on Fur_Recognition dataset and 6.41% on CIFAR-10 dataset. The algorithm in this paper can change the current situation that fur recognition relies on manual visual method to classify, and can lay foundation for improving the efficiency of fur production technology.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 2002년도 추계학술대회 논문집
• /
• pp.345-350
• /
• 2002

In this study, researchers classifying the artificial flaws in semiconductor packages are performed by pattern recognition technology. For this purposes, image pattern recognition package including the user made software was developed and total procedure including ultrasonic image acquisition, equalization filtration, binary process, edge detection and classifier design is treated by Backpropagation Neural Network. Specially, it is compared with various weights of Backpropagation Neural Network and it is compared with threshold level of edge detection in preprocessing method fur entrance into Multi-Layer Perceptron(Backpropagation Neural network). Also, the pattern recognition techniques is applied to the classification problem of defects in semiconductor packages as normal, crack, delamination. According to this results, it is possible to acquire the recognition rate of 100% for Backpropagation Neural Network.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 ITC-CSCC -3
• /
• pp.1523-1526
• /
• 2002

In this paper, we propose a method fur measuring the dimensions of an arbitrary object using geometric relationship between a perspective projection image and a rectangular parallelepiped model. For recognizing the vertexes of the rectangular parallelepiped surrounding an arbitrary object, the method adopts a strategy that derives the equations for vertex recognition from the geometrical relationships for image formation between 2D image and the rectangular parallelepiped model. extracts from 2D image with vertical view features (or junctions) of minimum quadrangle circumscribing an arbitrary shape object, and then recognizes vertexes from the features with the equations. Finally, the dimensions of the object are calculated from these results of vertex recognition. By the experimental results, it is demonstrated that this method is very effective to recognize the vertexes of the arbitrary objects.

• 한국정밀공학회지
• /
• 제22권9호
• /
• pp.85-92
• /
• 2005

This paper describes the recognition method of small-obstacles using a camera for a mobile robot in indoor environment. The technique of image processing using a camera has been widely used for an automaton of industrial system, an inspection of inferior goods, a lookout of an invader, and a vision sensor of intelligent robot. Mobile robot could meet small-obstacles such as a small plastic bottle of about 0.5 l in quantity, a small box of $7{\times}7{\times}7cm^3$ in volume, and so on in its designated path, and could be disturbed by them in the locomotion of a mobile robot. So, it is necessary to research on the recognition of small-obstacles using a camera and program. In this paper, 2-D image processing algorism and method fur recognition of small-obstacles using a camera for a mobile robot in indoor environment was developed. The characteristic test of the developed program to confirm the recognition of small-obstacles was performed. It is shown that the developed program could judge the size and the position of small-obstacles accurately.

• KIEE International Transaction on Systems and Control
• /
• 제12D권1호
• /
• pp.27-32
• /
• 2002

In this paper, we would study the applicability of neural networks to the recognition process of Korean stenographic character image, applying the classification function, which is the greatest merit of those of neural networks applied to the various parts so far, to the stenographic character recognition, relatively simple classification work. Korean stenographic recognition algorithms, which recognize the characters by using some methods, have a quantitative problem that despite the simplicity of the structure, a lot of basic characters are impossible to classify into a type. They also have qualitative one that It Is not easy to classify characters fur the delicacy of the character farms. Even though this is the result of experiment under the limited environment of the basic characters, this shows the possibility that the stenographic characters can be recolonized effectively by neural network system. In this system, we got 90.86% recognition rate as an average.

• 제어로봇시스템학회논문지
• /
• 제10권1호
• /
• pp.66-72
• /
• 2004

To assess tunnel safety, cracks in tunnel lining are measured by inspectors, who observe cracks with their naked eyes and record them. But manual inspection is slow, and measured crack data is subjective. Therefore, this study proposes inspection system fur measuring cracks in tunnel lining and providing objective crack data to be used in safety assessment. The system consists of On-vehicle system and Laboratory system. On-Vehicle system acquires image data with line CCD camera on scanning along the tunnel lining. Laboratory system extracts crack information from the acquired image using image processing. Measured crack information is crack thickness, length and orientation. To improve accuracy of crack recognition, the geometric properties and patterns of cracks in concrete structure were applied to image processing. The proposed system was verified with experiments in both laboratory environment and field environment such as subway tunnel.

• 지능정보연구
• /
• 제8권1호
• /
• pp.15-25
• /
• 2002

본 논문에서는 차량 번호판에서 추출된 문자영역의 DCT(Digital Cosine Transform) 계수와 LVQ(Learning Vector quantization) 신경회로망을 이용하여 상대적으로 간결한 구조로 잡음의 영향을 적게 받는 차량 번호판 인식 시스템을 제안하였다. 입력된 차량영상의 RGB칼라정보를 이용하여 번호판 영역을 추출하고 추출된 번호판의 히스토그램과 문자의 상대적 위치정보를 병합하여 문자영역을 추출하였다. 이렇게 추출된 문자영역의 명암도 영상에 DCT를 적용하여 얻은 특징 벡터를 LVQ신경회로망의 입력으로 사용하여 인식 과정을 수행한다. 본 논문의 실험과정에서는 다양한 환경에서 촬영된 109대의 자가용 차량영상에 대하여 제안된 시스템을 실험하였으며 상대적으로 높은 번호판 영역 추출율과 인식률을 보였다.

• 지능정보연구
• /
• 제12권1호
• /
• pp.1-15
• /
• 2006

우리나라의 주민등록증은 주소지, 주민등록번호, 얼굴사진, 지문 등 개인의 다양한 정보를 가진다. 현재의 플라스틱형 주민등록증은 위조 및 변조가 쉽고 그 수법이 날로 전문화 되어가고 있다. 따라서 육안으로 위조 및 변조 사실을 쉽게 확인하기가 어려워 사회적으로 문제를 일으키고 있다. 이에 본 논문에서는 개선된 ART2 기반 RBF 네트워크에 이용한 주민등록번호 인식과 얼굴 인증을 통한 주민등록증 자동 인식 방법을 제안한다. 제안된 방법은 주민등록증 영상으로부터 주민등록번호와 발행일을 추출하기 위하여 주민등록증 영상에 소벨 마스킹와 미디언 필터링을 적용한 후에 수평 스미어링을 적용하여 주민등록번호와 발행일 영역을 추출한다. 그리고 원영상에 대해 고주파 필터링을 적용하여 영상 전체를 이진화하고, 이진화된 영상에 CDM 마스크를 적용하여 주민등록번호와 발행일 코드를 복원한 다음, 검출된 각 영역에 대해 4-방향 윤곽선 추적 알고리즘을 적용하여 개별 문자를 추출한다. 추출된 주민등록번호 등의 개별 문자를 인식하기 위해 개선된 ART2 기반 RBF 네트워크를 제안하고 인식에 적용한다. 제안된 ART2 기반 RBF 네트워크는 학습 성능을 개선하기 위하여 중간층과 출력층의 학습에 퍼지 제어 기법을 적용하여 학습률을 동적으로 조정한다. 얼굴 인증은 템플릿 매칭 알고리즘을 이용하여 얼굴 템플릿 데이터베이스를 구축하고 주민등록증에서 추출된 얼굴 영역과의 유사도를 측정하여 주민등록증 얼굴 영역의 위조여부를 판별한다. 제안된 주민등록증 인식 방법의 성능을 평가하기 위해 원본 주민등록증 영상에 대해 얼굴 영역 위조, 노이즈추가, 대비 증감, 밝기 증감 그리고 영상 흐리기 등의 변형된 영상들을 생성하여 실험한 결과, 제안된 방법이 주민등록번호 인식 및 얼굴 인증에 있어서 우수한 성능이 있음을 확인하였다

• KSTLE International Journal
• /
• 제3권1호
• /
• pp.54-59
• /
• 2002

It can be effective on failure diagnosis of oil-lubricated tribological system to analyze operating conditions with morphological characteristics of wear debris in a lubricated machine. And it can be recognized that results are processed threshold images of wear debris. But it is needed to analyse and identify a morphology of wear debris in order to predict and estimate a operating condition of the lubricated machine. If the morphological characteristics of wear debris are identified by the computer image analysis and the neural network, it is possible to recognize the friction condition. In this study, wear debris in the lubricating oil are extracted from membrane filter (0.45 ${\mu}m$) and the quantitative value fur shape parameters of wear debris was calculated through the computer image processing. Four shape parameters were investigated and friction condition was recognized very well by the neural network.

• 한국산학기술학회논문지
• /
• 제10권3호
• /
• pp.554-559
• /
• 2009

본 논문에서는 인쇄체 숫자를 인식하기 위해 가중 원형 정합(weighted template matching) 방법을 제안한다. 원형 정합은 입력 영상 전체를 하나의 전역적인 특징으로 처리하는 데 반해, 제안된 가중 원형 정합은 패턴의 특징이 나타나는 국부적인 영역에 해밍 거리(Hamming distance)의 가중치를 두어 패턴 특징을 강조하여 숫자 패턴의 인식률을 높인다. 실험에서는 기존의 원형 정합을 사용했을 때, 오류 역전파 신경망을 사용했을 때와 가중 원형 정합을 사용했을 때의 혼돈 행렬(confusion matrix)을 각각 서로 비교한다. 실험 결과는 본 논문에서 제안한 방법에 의해 인쇄체 숫자의 인식률이 크게 향상된 것을 보인다.