• Journal of Sensor Science and Technology
• /
• v.11 no.2
• /
• pp.93-101
• /
• 2002

The quality of agricultural products is classified with various factors which are measured and determined by destructive and/or nondestructive method. NIR spectrum analysis method is used to determine internal qualities such as a brix and an acidity. CCD color camera is used to measure external quality like color and a size of fruit. Today, nondestructive methods are widely researched. The quality and the grade of fruit loaded into a cup automatically and measured in real time by camera and NIR system is determined by infernal and external factors. This paper proposes modified boundary tracking algorithm which detects the contour of fruit's color image and make chain code faster than conventional method. The chain code helps compute a size of fruit image and find multiple loading of a fruit in single cup or fruit between two cups. The designed classification system sorts a citrus at speed of 8 fruit/s, with evaluating a brix, an acidity and a size grade.

• Smart Media Journal
• /
• v.11 no.4
• /
• pp.46-55
• /
• 2022

It is very important to first detect and remove defective fruits with scratches or bruised areas in the automatic fruit quality screening system. This paper proposes a method of detecting defective areas in fruits using the latest artificial intelligence technology, the Yolo V4 deep learning model in order to overcome the limitations of the method of detecting fruit's defective areas using the existing image processing techniques. In this study, a total of 2,400 defective fruits, including 1,000 defective apples and 1,400 defective fruits with scratch or decayed areas, were learned using the Yolo V4 deep learning model and experiments were conducted to detect defective areas. As a result of the performance test, the precision of apples is 0.80, recall is 0.76, IoU is 69.92% and mAP is 65.27%. The precision of pears is 0.86, recall is 0.81, IoU is 70.54% and mAP is 68.75%. The method proposed in this study can dramatically improve the performance of the existing automatic fruit quality screening system by accurately selecting fruits with defective areas in real time rather than using the existing image processing techniques.

• Annual Conference on Human and Language Technology
• /
• 2008.10a
• /
• pp.177-182
• /
• 2008

본 논문에서는 질의응답시스템에서 응답 추출 대상 문서로 사용할 적절한 문서를 찾는 방법으로 기계 학습 기반의 문서 품질 평가 기법을 사용한다. 본 논문에서는 기존 연구와 달리 객관적인 정보를 많이 포함하고 있는 문서를 선별하는 목적으로 문서 품질 평가를 위한 유용한 자질들을 제안한다. 본 논문에서 정의한 정보성 자질은 정보의 양을 측정하는 자질과 정보의 객관성을 측정하는 자질로 구성된다. 실험 결과, 기존 문서 품질 평가 연구에서 주로 사용된 자질들만 사용한 경우와 새로운 자질들을 추가한 경우를 비교하였을 때, 1.5배 정도 높은 평균 정확률을 보였다. 제안하는 자질들 중에는 정보성 자질이 매우 유용한 자질이었고, 가독성 자질은 비교적 낮은 성능을 보였다. 문서의 여과 실험 결과, 96.4%의 재현율을 유지하면서 전체 문서 집합 중, 60%에 해당하는 저품질 문서를 여과할 수 있었다.

• Journal of the Korea Society of Computer and Information
• /
• v.13 no.6
• /
• pp.155-162
• /
• 2008

The quality of the fruits is measured by a lot of parameters. The grader of the fruits to measure the size of them is using the rotation drum method. Therefore when we classify the size of the fruits, they will be damaged. Also the optical grader used for estimating the degree of the saccharinity will incur high cost for it. In the proposed system, to select the characteristics of the fruits, three cameras are used. Because the information such as the volume and the degree of the maturity is used to classify the fruits, the degree of the saccharinity can't be estimated itself, but the information such as the color and the damage of the fruits can be estimated. Therefore, because we don't need the digital image with high resolution, we can develop the grader system of the fruit with low cost. To evaluate the performance of the proposed system, we compared it with the sight estimation and then we classified the sample. The result shows the accuracy of 96.7%.

• Land and Housing Review
• /
• v.1 no.1
• /
• pp.35-42
• /
• 2010

Most of the dredged sand generated from the sewage pipe maintenance project and the government's four-river project are disposed depending on abandonment and filling-up. This is caused by the lack of related recycling technology using dredged sand appropriately and high absorption rate and micro-particles of dredged sand producted from existing sand production system. Thus, this study carried out a quality assessment for the dredged sand produced through the optimum washing and sorting system supplementing problems of existing dredged sand production system as a part of research to examine performance of removing micro-particles and foreign substances. As a result of the assessment, the dredged sand produced through the cleaning and sorting system showed a wide quality improvement effect in absorption rate, 0.08 mm sieve pass amount, clay lump volume and organic impurity content, and it turned out to satisfy both the quality standards of this study, KS F 2573(recycled aggregate for concrete) and KS F 2526(aggregate for concrete) so it could be confirmed that it would be able to be used as an aggregate for concrete in the future.

• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.475-477
• /
• 2019

문장 부호란, 글에서 문장의 구조를 잘 드러내거나 글쓴이의 의도를 쉽게 전달하기 위하여 사용되는 부호들로, 따옴표나 쉼표, 마침표 등이 있다. 대화 시스템과 같이 컴퓨터가 생성해 낸 문장을 인간이 이해해야 하는 경우나 음성 인식(Speech-To-Text) 결과물의 품질을 향상시키기 위해서는, 문장 부호의 올바른 삽입이 필요하다. 본 논문에서는 이를 수행하는 딥 러닝 기반 모델을 훈련할 때 필요로 하는 한국어 말뭉치를 구축한 내용을 소개한다. 이 말뭉치는 대한민국정부에서 장관급 이상이 발언한 각종 연설문에서 적절한 기준을 통해 선별된 고품질의 문장으로 구성되어 있다. 문장의 총 개수는 126,795개이고 1,633,817개의 단어들(조사는 합쳐서 한 단어로 계산한다)로 구성되어 있다. 마침표와 쉼표는 각각 121,256개, 67,097개씩이다.

• Journal of Korean Society of Forest Science
• /
• v.110 no.2
• /
• pp.217-232
• /
• 2021

This study calculated the productivity and cost of extraction and processing of logging residues by cut-to-length (CTL) and whole-tree (WT) harvesting methods. In addition, the comparative analysis of the characteristics of wood chip fuel to examine whether it was suitable for the fuel conditions of the energy facility. In the harvesting and processing system to produce the wood chips of logging residues the system productivity and cost of the CTL harvesting system were 1.6 Gwt/SMH and 89,865 won/Gwt, respectively. The productivity and cost of the WT harvesting system were 2.9 Gwt/SMH and 72,974 won/Gwt, respectively. The WT harvesting productivity increased 1.3times while harvesting cost decreased by 18.7% compared to the CTL harvesting system. The logging residues of wood chips were not suitable for CTL wood chips based on International Organization for Standardization (ISO 17225-4:2021) and South Korea standard (NIFoS, 2020), but the quality (A2, Second class) was improved through screening operation. The WT-unscreened wood chips conformed to NIFoS standard (second class) and did not conform to ISO but were improved through screening operation (Second class). In addition to the energy facility in plant A, all wood chips except CTL-unscreened wood chips were available through drying processing. The WT-unscreened wood chips were the lowest at 99,408 won/Gwt. Plants B, C, and D had higher moisture content than plant A, so WT-unscreened wood chips without drying processing were the lowest at 57,204 won/Gwt. Therefore, the production of logging residues should improve with operation methods that improve the quality of wood chips required for applying the variable biomass and energy facility.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 2017.04a
• /
• pp.141-141
• /
• 2017

장미과(Rosaceae)에 속하는 딸기(Fragaria ananassa Duch.)는 비타민 C가 풍부하고 독특한 향기를 갖는 과채류로서 겨울에서 봄까지의 기간 동안 대부분 생식으로 소비되고 있다. 국내에서 재배되는 품종으로는 설향, 매향, 장희 등이 있으며 품종에 따라 성분과 함량이 다양하지만 일반적으로 유기산이 많아서 신맛과 단맛이 조화로운 특징이 있다. 소비자들이 딸기를 구입할 때 딸기가 포장된 상자에 모양이 일정하고 붉은 색상이 선명한 딸기에 호감을 갖게 된다. 딸기는 품종에 따라 기준이 되는 모양이 다르기 때문에 숙련된 선별사에 의해서 대부분 육안으로 선별되고 있는 실정이다. 하지만 개인적인 선별 능력의 차이와 주관적인 판단으로 인해 규격을 벗어난 딸기가 혼입되어 전체적인 품질 등급을 떨어뜨리는 경우가 종종 발생하기도 한다. 따라서 본 연구에서는 품종별로 기준이 되는 표준 형상과 비정상적인 모양의 기형 딸기를 객관적으로 판별하여 선별할 수 있는 영상 시스템을 구축하기 위해 수행되었으며 표준이 되는 딸기의 3차원 형상을 구축하기 위해 2차원 레이저 변위 센서를 이용하여 딸기의 입체 영상을 구축하고자 하였다. 실험을 위해 사용된 딸기는 시중에서 구입한 설향 품종이었으며 2차원 레이저 변위 센서는 라인 스캔 방식으로 1회 프로파일 스캔에 1,280개의 데이터 포인터를 획득할 수 있으며 분해능은 0.095~0.17 mm이었다. 상부에 부착된 2차원 레이저 변위 센서와 하부에 놓인 딸기의 거리는 100 mm였다. 획득한 딸기의 2차원 영상은 높이 차이를 이용하여 색상 농도로 표현하였으며 이 영상을 다시 3차원 영상으로 구축하였다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.109-114
• /
• 2022

질의 생성 모델은 스마트 스피커, 챗봇, QA 시스템, 기계 독해 등 다양한 서비스에 사용되고 있다. 모델을 다양한 서비스에 잘 적용하기 위해서는 사용자들의 실제 질의 특성을 반영한 자연스러운 질의를 만드는 것이 중요하다. 본 논문에서는 사용자 질의 특성을 반영한 간결하고 자연스러운 질의 자동 생성 모델을 소개한다. 제안 모델은 topic 키워드를 통해 모델에게 생성 자유도를 주었으며, 키워드형 질의→자연어 질의→응답으로 연결되는 chain-of-thought 형태의 다중 출력 구조를 통해 인과관계를 고려한 결과를 만들도록 했다. 최종적으로 MRC 필터링과 일관성 필터링을 통해 고품질 질의를 선별했다. 베이스라인 모델과 비교해 제안 모델은 질의의 유효성을 크게 높일 수 있었다.

• Smart Media Journal
• /
• v.12 no.1
• /
• pp.9-16
• /
• 2023

Recently, a convolutional neural network (CNN) based system is being developed to overcome the limitations of human resources in the apple quality classification of farmhouse. However, since convolutional neural networks receive only images of the same size, preprocessing such as sampling may be required, and in the case of oversampling, information loss of the original image such as image quality degradation and blurring occurs. In this paper, in order to minimize the above problem, to generate a image patch based graph of an original image and propose a random walk-based positional encoding method to apply the graph transformer model. The above method continuously learns the position embedding information of patches which don't have a positional information based on the random walk algorithm, and finds the optimal graph structure by aggregating useful node information through the self-attention technique of graph transformer model. Therefore, it is robust and shows good performance even in a new graph structure of random node order and an arbitrary graph structure according to the location of an object in an image. As a result, when experimented with 5 apple quality datasets, the learning accuracy was higher than other GNN models by a minimum of 1.3% to a maximum of 4.7%, and the number of parameters was 3.59M, which was about 15% less than the 23.52M of the ResNet18 model. Therefore, it shows fast reasoning speed according to the reduction of the amount of computation and proves the effect.