Abstract
(1)農業生產與統計視覺化之研究：輔助政府蒐集正確且具時效性之農業統計資料如主力農家所得、農業勞動力等農業統計調查。整合各單位價量資料，提供農產品進出口貿易、價量與生產量值視覺化查詢介面，協助決策者快速掌握農產品資訊。(2)運輸車輛軌跡查核管理系統應用計畫：開發共通即時傳輸功能，提供不同廠商之車輛軌跡管理系統即時轉抛即時運輸車輛串流資訊，再運用農委會既有畜牧場地理資訊，配合農委會防檢疫需要，開發即時運輸車輛追蹤、查核與告警等相關功能，以及歷史軌跡處理支援以場追車，以車查場等功能，支援非洲豬瘟防疫指揮管制任務。(3)深度學習技術輔助遙測影像判釋研究計畫：彙集整理多年度航照影像圖資資訊，作為研發及驗證使用。以深度學習技術達到快速有效的農作物坵塊之自動標示與判釋，利用Unet-VGG16作為進行影像分割的深度學習方法，搭配植生指標NDVI及加入水稻物候期的標示，可以提昇整體判釋準確度，透過使用者友善的操作介面，讓一般操作者皆利用所開發的AI工具，在地理圖資系統中快速、準確地標識水稻的栽種區域。(4)ESRI ArcGIS地理資訊應用軟體大量授權暨教育訓練案：提供農委會暨所屬機關既有ESRI ArcGIS地理資訊應用軟體數量升級維護服務，提供原廠使用授權；製作6門數位課程及辦理教育訓練，包含實體課程總計至少15場、線上QA總計至少20場。&zwsp;&zwsp;&zwsp;&zwsp;&zwsp;&zwsp;&zwsp;(5)農業地理資訊系統資訊及運算資源共構計畫：提供雲端運算的快速佈署、彈性擴充以及GPU運算等等各類運算情境的使用並支援各項基礎資料建置所需的資源，協助發展農業GIS協作，以強化農委會農業空間管理應用，解決農委會所屬機關中資源不均與不足的問題，加速各種資料的共享運用，以期利用於各項農業生產資訊的掌握，並整合農業領域各項應用，最終達到提升農業施政效率。(6)整合作物現地調查判釋計畫：透過作物調查行動應用程式(APP)，整合支援農委會暨所屬機關各類型作物現地調查，累計收集農田現地照片約200萬張，完成現地調查照片有效判釋80萬筆，判釋正確率應達95%以上。(7)農業土地資源總盤點，調查高污染風險區農地重金屬，確認農地質與量計畫：整合全台農業區與農村的地籍圖並套疊土壤、海拔、土地利用、高解析度衛星圖像、數值地形模型（DTM）和地質資訊，製作農業土地利用系統類型之土地面積的地圖。並配合農委會資訊中心的農業空間資料標準化整理空間座標，並與地籍圖及具地理參考座標的衛星影像、地形、數值地形模型等，建立農業土地圖像與地籍結合的資料庫。 The Study on Agricultural Production Statistics and Data VisualizationChallenges faced in Agricultural development today are complex and diverse, including the problems raising in productions, demands, workforce, water usages, and subsidies. This project aims to integrate the farming, livestock, and welfare databases for agricultural policy assessments and decisions. The integrated databases can be applied for: (1) Provide the administrative sources that increases the accuracy of agricultural surveys, e.g. the Major Farm Household Income Survey and Agricultural Labor Resource Survey. (2) Integrate the international trade, price and yield data to provide a visual inquiry interface for agricultural product information.
The goals of this project will be accomplished from three perspectives: (1) Import, clean and integrate 20 millions data each year from farming, livestock, and welfare databases to provide the governmental data as supplement file that increase the accuracy and efficiency of agricultural survey (e.g. the Major Farm Household Income Survey, Agricultural Labor Resource Survey.). (2) Expand the important agricultural product price and yield information platform and agricultural product trade statistics information platform to provide convenient visual query functions. (3) Develop the agricultural production statistics information platform to provide convenient interactive visual query functions. The outcome of this project aims to provide the information for future agricultural policy assessments and decisions by a visualization interface.
Transportation Vehicle Track Check Management System Application PlanDue to the rapid spread of the African swine fever epidemic in mainland China, my country immediately initiated the strengthening of border epidemic prevention and control work to prevent possible epidemic intrusion. On December 18, 107, it deployed in advance to establish a response center to comprehensively analyze the possible intrusion gaps and immediately To prevent congestion, at the same time strengthen the inventory of various epidemic prevention and preparation work with local governments. The response center emphasized that in addition to border control measures for the prevention and control of African swine fever, the transformation of food waste on pig farms and the installation of GPS monitoring on pig transportation vehicles are used to facilitate scientific and technological law enforcement to "track and chase vehicles on the field" to implement epidemic tracking and Control, ensure that the epidemic prevention is drip-proof.The first phase (108 years) of this project is to cooperate with the "Science and Technology Law Enforcement Real-time Tracking Flow Direction, Comprehensive Prevention of African Swine Fever", to develop a common real-time transmission function, and provide real-time transfer and real-time transportation vehicle streaming information of different manufacturers' vehicle trajectory management systems , And then use the geographic information of the existing livestock farms of the Council of Agriculture to meet the needs of the Council of Agriculture for prevention and quarantine, develop real-time transportation vehicle tracking, inspection and warning and other related functions, as well as historical trajectory processing to support vehicle tracking on the field and vehicle inspection of the factory. , To support African swine fever prevention and control missions.In this period (year 109), additional repairs will be carried out on the basis of the above to achieve the establishment of an early warning mechanism, the preparation of domestic breeding management and slaughter hygiene and transportation management and other operations to respond to various emergencies as necessary measures, and to plan epidemic prevention Control areas to prevent the spread of the epidemic and achieve the goal of integrated use of epidemic prevention resources.
Assisting agricultural image interpretation with deep learning techniquesIn the present, the development of agriculture and the promotion of policies often rely on artificial intelligence. Through machine learning or deep learning image recognition technology, it can be applied to the interpretation of crop aerial images, which can improve the interpretation efficiency and reduce the interpretation cost. Because of this, the Council of Agriculture (COA) commissioned research and academic institutions to develop AI aerial image interpretation model. To verify these models’ benefits, this plan will collect the multi-year aerial image information for development and verification. This project focuses on the issue of automatic rice field segmentation and labeling at different rice growth stages. We will develop a deep learning model on top of the Unet-VGG16 architecture for semantic segmentation of rice fields at each rice growth stage. Besides, we will develop deep learning models based on several vegetation indices such as NDVI and integrate deep learning models of the different stages of rice into the geographic information system (GIS) platform as modules to improve the overall interpretation accuracy and efficiency through a user-friendly operation interface. The interface allows general users to apply the developed AI tools to quickly and accurately identify various crop cultivation areas in the GIS, such as rice.
Provide Extensive Authorization of ESRI ArcGIS with Training of Geographic Information System (GIS) Application SoftwareFor efficient land management of the agricultural, forestry, fishery, and husbandry, The Council of Agriculture (COA) Associates of Executive Yuan, make use of GIS system software tools to support and execute their multi-task operations. By excellent computing capacity, analysis functionality, and long-term promotion of ESRI GIS software to generate the best performance, ESRI ArcGIS becomes the main GIS software tools to execute its related tasks in the COA Executive Yuan. As planned, there are four main items to be executed as follows:(1)Provide upgrade and maintenance services for COA, Executive Yuan and its affiliated institutions’ existing application software licenses of ESRI ArcGIS, as listed below: 593 sets of ArcGIS Desktop, 170 sets of ArcGIS Desktop Extensions, 22 sets of ArcGIS Enterprise, and a set of Server Roles.(2)Provide unlimited trial licenses of ESRI ArcGIS without extra payment during the contract period of the project. Includes ArcGIS Desktop(ArcMap and ArcGIS Pro), ArcGIS 3D Analyst(Extension), ArcGIS Network Analyst(Extension), ArcGIS Spatial Analyst(Extension), ArcGIS Enterprise, ArcGIS Server, ArcGIS Image Server, ArcGIS GeoEvent Server and ArcGIS GeoAnalytics Server.(3)Produce 6 digital courses, offer at least 15 days of training and 20 sessions of Q&A.(4)Provide consultation service lines and nationwide (divided into 4 regions: North, Central, South, and East) technical services such as ArcGIS software upgrade, installation, configuration settings, etc.
The Computing Service for the Project of the Integrated Applications with Agricultural Geographic Information System(AGIS)Main objective of this project is to integrate the crucial hardware and computing facilities for the establishment of Agricultural Geographic Information System (AGIS). The AGIS is built on the cloud computing of NCHC which is easy to access and available on demand including GPU deployment, VCS and cloud storage capacity expansion etc. This AGIS platform will benefits a lot from such infrastructures of NCHC. For example, it can improve the data sharing (e.g. agricultural imagery management,) between administrative governances in central and local governments which will level-up the efficiency of routine work of crop productions management. Major tasks of this project includes: 1. The hardware facilities for AGIS platform
The projects of the crop site-survey integrating with the interpretation of the site photosCombining with the existing "Agricultural Field Survey Operation Management System" of the Council of Agriculture, through the Crop Survey Action App, this project integrates and supports the Council of Agriculture and its affiliated agencies for field surveys of various types of crops, and has collected about 2 million farmland on-site photos. Because each government unit (agency) has different purposes for conducting on-site surveys, and the methods of crop classification of survey results are different, there is an obstacle to sharing the results of on-site surveys across units (agency).Therefore, this project adopts the background manual interpretation interface of the "Agricultural Field Survey Operation Management System", and artificial intelligence (AI) interpretation modules for rice, banana and pineapple categories. Regarding the newly-added and existing field photos of the Council of Agriculture in 2021, the crop classification based on the Agricultural Survey of the Agriculture and Food Agency (about 250 crops in Taiwan) is used as the basis for crop classification. The effective interpretation of 800,000 photos is to be completed on the spot investigation. The accuracy rate of interpretation should reach more than 95%. It is to conduct the development of expert, UAV and AI integrated crop interpretation module. In addition to providing cross-unit (agency) sharing of on-site survey results, the results of this project also provide personnel retraining of existing AI interpretation modules through expert interpretation results, as well as the addition of multiple crop AI module operations. Furthermore, the project handles a total of 800 kilometers and 40,000 photos of farmland street scenes adjacent to the farmland parcels, collected and interpreted to enhance the ability of integrated interpretation.
Agricultural land quality and quantity inventor Based on the preliminary inventory of agricultural land completed in 2016-2017, the project continuously update the agricultural land data, such as the distribution of land use and area, to obtain map resources before the base date: including cadastre, ultra-high resolution satellite photos, aerial photography images, important agricultural production areas, agricultural facilities, etc. The project will priority complete the classification of 4/5 classified land cover / utilization layer and digitalization of agricultural land data in the vicinity of the adjacent threatened areas of agricultural land data with field surveys. On the other hand, the cadastral maps of agricultural areas and rural areas in Taiwan will be integrated, and soil, altitude, land use, high-resolution satellite imagery, numerical topographic model (DTM) and geological information will be integrated to produce one of the types of agricultural land use systems. Coordinate with the agricultural space data of the COA Information Center to standardize the space coordinates, and establish a database of agricultural land images and cadastre with the cadastral map and satellite imagery, topographical and numerical terrain models with georeferenced coordinates. The database of the two crops of the plain area will be completed in 2021, covered with 1.08 million hectares. Transfer the established technology and system of remote measurement and spatial information survey to 3 county and city governments and central management units for inventory and estimation of agricultural land area and farming status. Organize and analyze the results of the detailed survey of farmland soil in 30,000 hectares of high-risk areas of heavy metal pollution, understand the changing trend of farmland quality, prevent heavy metal pollution in the food chain, and incorporate environmental spatial data in response to the needs of Global GAP farmland environmental assessment.
