표제지
목차
요약문 5
SUMMARY 6
제1장 연구개발과제의 개요 11
제1절 연구개발의 목적 및 필요성 11
1. 연구목적 11
2. 연구의 필요성 11
제2장 국내외 기술개발 현황 13
1. 국내 기술개발 현황 13
2. 국외 기술개발 현황 14
제3장 연구개발 수행내용 및 결과 15
제1절 해양수산정보 활용 빅데이터 분석 15
1. 해양수산정보 활용을 위한 AI 적용기술 개발 및 시험적용 15
2. 위성 관측자료 활용 해양정보예측 17
3. 현장 부이 관측자료 활용 해양정보예측 22
4. 빅데이터 분석 서비스(수요자 맞춤형 자료) 26
제2절 경험 정보 분석 29
1. 해역별 어업생산 동향 어업인 청취 조사 29
2. 어업 경험 정보 데이터화를 위한 정량지표 개발 35
제3절 데이터 표준화 구축 36
1. 표준화 목적 및 대상 데이터의 특징 36
2. 표준화 분석 및 정립 37
제4절 수산과학 빅데이터 시스템 구축 44
1. 빅데이터 시스템 표준화 기술개발 44
2. 빅데이터 관리 정책 수립 48
3. 빅데이터 처리 성능개선연구 50
4. 빅데이터 플랫폼 통합제어 시스템 개발 51
5. 수집자료 분산 저장 및 처리성능 최적화 53
6. 빅데이터 플랫폼 시스템 구현방안 55
제5절 빅데이터와 인공지능을 활용한 최적 예측 수산정보 생산방안 60
1. 생태계변동 예측모델 정확도 향상 방안 60
2. 한국형 연근해 생태계변동 예측시스템 연계 67
제6절 비정형 자료 온톨리지 구축 및 수산정보 생산 기술 연구 70
1. 비정형 자료 온톨로지 구축 및 수산정보 생산 기술 연구 70
제7절 맞춤형 수산정보 가시화 방안 연구 77
1. 맞춤 수산정보 통합 DB 관리시스템 77
2. 대국민 서비스 시스템 81
3. 모바일 서비스 방안 연구 84
제4장 목표 달성도 및 관련 분야에의 기여도 86
1. 목표 달성도 86
2. 대표 성과 및 기여도 86
제5장 연구개발 결과의 활용계획 89
제6장 참고문헌 90
제7장 부록 93
판권기 94
Table 1. NOAA satellite SST observation data specifications 18
Table 2. Prediction model design of weekly SST 19
Table 3. Monthly prediction performance statistic for weekly mean SST in 2019 20
Table 4. Seasonal prediction performance statistic for weekly mean SST in 2019 20
Table 5. Prediction model design of monthly satellite SST 21
Table 6. Prediction performance statistic for monthly mean SST in 2020 22
Table 7. Prediction model design of buoy observation data 25
Table 8. Prediction performance statistic for buoy observation SST data in 2019 26
Table 9. Percentage of main target fish species for fishing activities in the research area 33
Table 10. Percentage of fishing gear for fishing activities in the research area 34
Table 11. Data type and items of NIFS database (2018. 11. 01.) 37
Table 12. Standardization for spatial-temporal information of in-situ data 38
Table 13. Standardization for Biomass, Nutrients and Chlorophyll-a values 38
Table 14. Current status of variable standardization for ecosystem model results (part) 39
Table 15. Standardization for in-situ observation data 39
Table 16. Standardization for Satellite image data 41
Table 17. Standardization code for Satellite image file naming 41
Table 18. Standardization for Satellite observation information 42
Table 19. Standardization for data processing level 42
Table 20. Standardization for satellite products 43
Table 21. Standardization for satellite products 43
Table 22. Standardization for algorithm of products 44
Table 23. Metadata categories of Fisheries research data 49
Table 24. Metadata categories of research report 49
Table 25. Types of cloud-based system performance tests 50
Table 26. Lookup module of structured data using PySpark 56
Table 27. Module of Apache Hive distributed storage 57
Table 28. Input data for chlorophyll-a prediction 62
Table 29. Prediction model design of chlorophyll-a 62
Table 30. Performance of chlorophyll-a prediction (2018) 63
Table 31. Prediction model design of primary productivity 64
Table 32. 7-day prediction model design of satellite SST 66
Table 33. Statistics of 7-day satellite SST prediction results 66
Table 34. Pre-processing, model run, post-processing module 68
Table 35. Static result of deepfillv2 and hypergraphs model (2019) 70
Table 36. Information of data collection from news and SNS (blog) 71
Table 37. Difference between mobile and PC web site 85
Fig. 1. Concept map of personalized fisheries information service 11
Fig. 2. Satellite observation path data (A) and orthorectified satellite data (B) and weekly satellite Sea Surface Temperature (SST) (C) 15
Fig. 3. The data to be analyzed and the applicable techniques of AI 15
Fig. 4. Results of monthly satellite image restoration through GAN. January (A), May (B), September (C) 17
Fig. 5. Example of weekly (A) and monthly (B) mean SST derived from NOAA 18
Fig. 6. ConvLSTM algorithm internal structure (A), prediction structure (B) 18
Fig. 7. Example of weekly mean SST 19
Fig. 8. Example of weekly mean SST calculation 19
Fig. 9. Result of weekly satellite SST AI prediction week 11 of 2019 (A), week 33 of 2019 (B) and groundtruth week 11 of 2019 (C), ground truth week 33 of 2019 (D) 20
Fig. 10. Example of monthly mean SST calculation 21
Fig. 11. Result of monthly mean SST AI prediction May 2020 (A), ground truth May 2020 (B) and September 2020 (C), ground truth September 2020 (D) 22
Fig. 12. LSTM structure (A) and cell gate (B) 23
Fig. 13. Timeseries of observation data Seosan (A), Boryeong (B), Yeosu (C), Gijang (D), Guryongpo (E), Yeongdeok (F), Samcheok (G) 23
Fig. 14. EMD results of Seosan, IMF-1 (A), IMF-2 (B), IMF-3 (C), IMF-4 (D), IMF-6 (E), Trend (F) 24
Fig. 15. Prediction results of Seosan (A), Boryeong (B), Yeosu (C), Gijang (D), Guryongpo (E), Yeongdeok (F), Samcheok (G) in 2019 25
Fig. 16. Flowchart of keyword-based analysis (A), text mining analysis main page (B) 27
Fig. 17. A schematic diagram of Unstructured data collection 27
Fig. 18. Example of the result of word tokenization 28
Fig. 19. 'Red tide' keyword search result of textmining platform, main page (A), images (B), list of documents (C), LDA analysis (D), network analysis (E) 29
Fig. 20. Experience information collection interview (A) and record sheet (B) 30
Fig. 21. Distribution pattern by study area (A) and respondents by age (B) 30
Fig. 22. Main target species of fishing (A) and fishing gear used in fisheries (B) 31
Fig. 23. Fishing gear installation and standard of use (A), Basis for decisions on fishing (B) 31
Fig. 24. Distribution pattern by study area (A) and respondents by age (B) 32
Fig. 25. Main target species of fishing (A), fishing gear used in fisheries (B) 32
Fig. 26. Determinants of fishing (A) and fishing gear installation (B) and source of experience information acquisition (C) 35
Fig. 27. Database of field investigation results (part) 36
Fig. 28. Processing solution of grid data 46
Fig. 29. Processing solutions of structured data real-time (A), batch (B) 47
Fig. 30. Cloud virtual resource performance measurement test flowchart 50
Fig. 31. Apache Hadoop ecosystem 51
Fig. 32. Integrated management of Apache Ambari big data solutions. Dashboard (A), services (B), hosts (C~D), alerts (E), admin (F) 52
Fig. 33. Big data formats comparison 53
Fig. 34. Relationship between cores, memory, and executor 54
Fig. 35. Schematic diagram of structured data analysis using Apache Spark 55
Fig. 36. Result of MYOCEAN using Jupyter-lab 56
Fig. 37. Converting NetCDF MYOCEAN data to parquet 58
Fig. 38. Physical-geochemistry-ecological model linkage diagram 61
Fig. 39. Multivariate multidimensional neural network diagram (A) and ConvLSTM structure of multivariate input data (B) 61
Fig. 40. Input data collected for chlorophyll-a prediction, chlorophyll-a (A), PAR (B), SST4 (C), Z_eu (D) 62
Fig. 41. Chlorophyll-a prediction results using ConvLSTM, AI (A), (B), ground Truth (C), (D) 63
Fig. 42. Chlorophyll-a (A), PAR (B), SST4(C), Kd490 (D) from VIIRS-SNPP satellite and in-situ PP (E) 64
Fig. 43. Prediction result of primary productivity using DNN (A) and R2 score (B), Comparison result of observation, DNN and K&I algorithm (C) 65
Fig. 44. 7-day prediction digram of satellite SST 65
Fig. 45. System integration diagram of inpainting AI model and Ecosystem variation prediction system (A), and AI interpolation selection screen (B) 67
Fig. 46. Program manual for interpolation of missing satellite data using AI (B) 68
Fig. 47. Diagram comparing Deepfillv2 and Hypergraphs model 69
Fig. 48. Input data of January (A), June (E) 2019 and ground truth data of January (B), June (F) 2019 and deepfillv2 result of January (C), June (G), hypergraphs... 70
Fig. 49. Example of ontology concepts and components 71
Fig. 50. Natural language processing and data analysis of unstructured data. on ontology 73
Fig. 51. Example of sentence tokenization in 'article' class 73
Fig. 52. Diagram of unstructured data analysis 74
Fig. 53. Frequency of word appearance (A), network analysis (B), emotional analysis (C) and topic modeling (LDA) analysis (D) 76
Fig. 54. AFIS user screen 78
Fig. 55. Excel management on web system menu (A) and process to check excel file data format version (B) 79
Fig. 56. Searching observation code 79
Fig. 57. A dashboard that shows the status of data construction over past 10 years 80
Fig. 58. A dashboard indicating statistical information on marine environment (A) and fisheries resources (B) 80
Fig. 59. Main page corresponding to forecast and breaking news service such as forecast water temperature based on satellite data 81
Fig. 60. A weekly water temperature prediction service based on real-time observation bouy data using AI 82
Fig. 61. SST prediction service using machine-learning based on satellite SST images 83
Fig. 62. Opinion registration window for online users. Improvement opinion registration screen of the integrated DB platform (A), Diagram of the manual for... 84