• International journal of advanced smart convergence
• /
• 제7권2호
• /
• pp.95-100
• /
• 2018

In this paper, we propose a binarized multi-scale module to accelerate the speed of the pose estimating deep neural network. Recently, deep learning is also used for fine-tuned tasks such as pose estimation. One of the best performing pose estimation methods is based on the usage of two neural networks where one computes the heat maps of the body parts and the other computes the part affinity fields between the body parts. However, the convolution filtering with a large kernel filter takes much time in this model. To accelerate the speed in this model, we propose to change the large kernel filters with binarized multi-scale modules. The large receptive field is captured by the multi-scale structure which also prevents the dropdown of the accuracy in the binarized module. The computation cost and number of parameters becomes small which results in increased speed performance.

• 지능정보연구
• /
• 제9권2호
• /
• pp.1-18
• /
• 2003

The automatic recognition of transport containers using image processing is very hard because of the irregular size and position of identifiers, diverse colors of background and identifiers, and the impaired shapes of identifiers caused by container damages and the bent surface of container, etc. We proposed and evaluated the novel recognition algorithm of container identifiers that overcomes effectively the hardness and recognizes identifiers from container images captured in the various environments. The proposed algorithm, first, extracts the area including only all identifiers from container images by using CANNY masking and bi-directional histogram method. The extracted identifier area is binarized by the fuzzy binarization method newly proposed in this paper and by applying contour tracking method to the binarized area, container identifiers which are targets of recognition are extracted. We proposed and applied the ART2-based RBF network for recognition of container identifiers. The results of experiment for performance evaluation on the real container images showed that the proposed algorithm has more improved performance in the extraction and recognition of container identifiers than the previous algorithms.

• 센서학회지
• /
• 제30권1호
• /
• pp.20-24
• /
• 2021

The electrocardiogram (ECG) is one of the most extensively employed signals used to diagnose and predict cardiovascular diseases (CVDs). In recent years, several deep learning (DL) models have been proposed to improve detection accuracy. Among these, deep neural networks (DNNs) are the most popular, wherein the features are extracted automatically. Despite the increment in classification accuracy, DL models require exorbitant computational resources and power. This causes the mapping of DNNs to be slow; in addition, the mapping is challenging for a wearable device. Embedded systems have constrained power and memory resources. Therefore full-precision DNNs are not easily deployable on devices. To make the neural network faster and more power-efficient, spiking neural networks (SNNs) have been introduced for fewer operations and less complex hardware resources. However, the conventional SNN has low accuracy and high computational cost. Therefore, this paper proposes a new binarized SNN which modifies the synaptic weights of SNN constraining it to be binary (+1 and -1). In the simulation results, this paper compares the DL models and SNNs and evaluates which model is optimal for ECG classification. Although there is a slight compromise in accuracy, the latter proves to be energy-efficient.

• 한국지능시스템학회논문지
• /
• 제14권3호
• /
• pp.349-356
• /
• 2004

The automatic recognition of transport containers using image processing is very hard because of the irregular size and position of identifiers, diverse colors of background and identifiers, and the impaired shapes of identifiers caused by container damages and the bent surface of container, etc. In this paper we propose and evaluate a novel recognition algorithm for container identifiers that effectively overcomes these difficulties and recognizes identifiers from container images captured in various environments. The proposed algorithm, first, extracts the area containing only the identifiers from container images by using CANNY masking and bi-directional histogram method. The extracted identifier area is binarized by the fuzzy binarization method newly proposed in this paper. Then a contour tracking method is applied to the binarized area in order to extract the container identifiers which are the target for recognition. In this paper we also propose and apply a novel ART2-based hybrid network for recognition of container identifiers. The results of experiment for performance evaluation on the real container images showed that the proposed algorithm performs better for extraction and recognition of container identifiers compared to conventional algorithms.

• 한국산학기술학회:학술대회논문집
• /
• 한국산학기술학회 2003년도 Proceeding
• /
• pp.88-95
• /
• 2003

The automatic recognition of transport containers using image processing is very hard because of the irregular size and position of identifiers, diverse colors of background and identifiers, and the impaired shapes of identifiers caused by container damages and the bent surface of container, etc. We proposed and evaluated the novel recognition algorithm of container identifiers that overcomes effectively the hardness and recognizes identifiers from container images captured in the various environments. The proposed algorithm, first, extracts the area including only all identifiers from container images by using CANNY masking and bi-directional histogram method. The extracted identifier area is binarized by the fuzzy binarization method newly proposed in this paper and by applying contour tracking method to the binarized area, container identifiers which are targets of recognition are extracted. We proposed and applied the ART2-based RBF network for recognition of container identifiers. The results of experiment for performance evaluation on the real container images showed that the proposed algorithm has more improved performance in the extraction and recognition of container identifiers than the previous algorithms.

• 제어로봇시스템학회논문지
• /
• 제21권3호
• /
• pp.276-282
• /
• 2015

This paper proposes a SMD (surface mounting device) classification method for the PCB assembly inspection machines. The package types of SMD components should be classified to create the job program of the inspection machine. In order to reduce the creation time of job program, we developed the automatic classification algorithm for the SMD packages. We identified the chip-type packages by color and edge distribution of the images. The input images are transformed into the HSI color model, and the binarized histroms are extracted for H and S spaces. Also the edges are extracted from the binarized image, and quantized histograms are obtained for horizontal and vertical direction. The neural network is then applied to classify the package types from the histogram inputs. The experimental results are presented to verify the usefulness of the proposed method.

• 한국컴퓨터정보학회논문지
• /
• 제12권2호
• /
• pp.77-82
• /
• 2007

본 논문에서는 콘크리트 표면 균열의 방향성을 효율적으로 인식하기 위하여 영상처리 기법을 적용하여 균열을 자동으로 검출하고 개선된 Max-Min 신경망을 제안하여 균열의 방향성을 자동으로 인식하는 기법을 제안한다. 균열 영상에서 빛의 영향을 효율적으로 보정하기 위해 모폴로지 기법인 채움 연산을 적용하고 Sobel 마스크를 적용하여 균열의 에지를 추출한 후, 반복 이진화를 적용하여 균열 영상을 이진화한다. 이진화된 균열 영상에서 2차례에 걸쳐 잡음 제거 연산을 수행한 후, 균열을 추출한다. 본 논문에서는 Max-Min 신경망을 개선하여 추출된 균열의 방향성을 자동으로 인식한다. 개선된 Max-Min 신경망은 delta-bar-delta 알고리즘을 적용하여 학습률을 자동으로 조정한다. 실제 콘크리트 표면 균열 영상을 대상으로 실험한 결과, 개선된 Max-Min 신경망이 균열의 방향성 인식에 효율적임을 확인하였다.

• Journal of information and communication convergence engineering
• /
• 제6권2호
• /
• pp.213-217
• /
• 2008

This paper proposes a novel method for the recognition of passports based on the CDM(Conditional Dilation Morphology) masking and the ART2-based RBF neural networks. For the extraction of individual codes for recognizing, this paper targets code sequence blocks including individual codes by applying Sobel masking, horizontal smearing and a contour tracking algorithm on the passport image. Individual codes are recovered and extracted from the binarized areas by applying CDM masking and vertical smearing. This paper also proposes an ART2-based hybrid network that adapts the ART2 network for the middle layer. This network is applied to the recognition of individual codes. The experiment results showed that the proposed method has superior in performance in the recognition of passport.

• 한국항행학회논문지
• /
• 제25권5호
• /
• pp.426-432
• /
• 2021

CW (continuous wave) 도플러 레이다는 카메라와 달리 사생활 침해 문제를 해결할 수 있고, 비접촉 방식으로 신호를 얻을 수 있다는 장점이 있다. 따라서, 본 논문에서는 CW 도플러 레이다를 이용한 사람 행동 인식 시스템을 제안하고, 가속을 위한 하드웨어 설계 및 구현 결과를 제시한다. CW 도플러 레이다는 사람의 연속된 동작에 대한 신호를 측정한다. 이에, 동작 분류를 위한 단일 스펙트로그램을 얻기 위해 운동 동작의 횟수를 세는 기법을 제안하였다. 또한, 연산의 복잡도와 메모리 사용량을 최소화하기 위해 동작 분류에 BNN (binarized neural network)을 사용하였고, 검증 결과 94%의 정확도를 보임을 확인하였다. BNN의 복잡한 연산을 가속하기 위해 FPGA를 이용하여 BNN 가속기가 설계 및 구현되었다. 제안된 사람 행동 인식 시스템은 logic 7,673개, register 12,105개, combinational ALUT (adaptive look up table) 10,211개, block memory 18.7 Kb를 사용하여 구현되었으며, 성능 평가 결과 소프트웨어 구현 대비 연산 속도가 99.97% 향상되었다.

• 한국지능시스템학회논문지
• /
• 제14권2호
• /
• pp.222-227
• /
• 2004

Today, an automatic and accurate processing using computer is essential because of the rapid increase of travelers. The determination of forged passports plays an important role in the immigration control system. Hence, as the preprocessing phase for the determination of forged passports, this paper proposes a novel method for the recognition of passports based on the fuzzy binarization and the fuzzy RBF network. First, for the extraction of individual codes for recognizing, this paper targets code sequence blocks including individual codes by applying Sobel masking, horizontal smearing and a contour tracking algorithm on the passport image. Then the proposed method binarizes the extracted blocks using fuzzy binarization based on the trapezoid type membership function. Then, as the last step, individual codes are recovered and extracted from the binarized areas by applying CDM masking and vertical smearing. This paper also proposes an enhanced fuzzy RBF network that adapts the enhanced fuzzy ART network for the middle layer. This network is applied to the recognition of individual codes. The results of the experiments for performance evaluation on the real passport images showed that the proposed method has the better performance compared with other approaches.