Crowdsourcing For Inforniationalized Forecasting Of Agricultural Pests

Preventing and controling agricultural pests are key ways to maintain steady grain output and to increase the efficiency of grain production,as well as to improve grain quality.It is obviously that preventing and controlling agricultural pests are the most important supports for the development of modern agriculture.With the rapid development of the Internet(World Wide Web)and computer technology,Informationized forecasting method has become one of the important ways of preventing and controling agricultural pests in China.The existing informationized approaches employing knowledge bases or ontologies,such as mathematical statistics,expert system remote diagnosis et al.,can significantly improve the accuracy of agricultural pest forecasting.However,depending on domain experts to construct and to maintain knowledge bases(and ontologies)results in slow update,which further leads to the limitations of poor timeliness of information acquisition and incomplete condition.The main reasons are 1)the contradiction between small-scale domain experts and large-scale tasks leads to delay in updating the knowledge bases(and ontologies);and 2)the high cost of hiring domain experts reduces the update frequency of knowledge bases(and ontologies).Fortunately,Crowdsourcing has huge advantages in terms of hiring large-scale people with low costs.Therefore,this thesis focuses on the problem of crowdsourcing-based approach of forecasting agricultural pests.Our research contributions are as follows.(1)Information extraction of agricultural pests using crowdsourcing technology.Farmers first take photographs or videos of agricultural pests they discovered by using smart phones,and then upload them attached with a short text to systems.Since Image matching and short text analysis are computer-hard tasks,it is natural for us to employ human intelligence.Hiring domain experts can achieve high accuracy,however,the cycle is long and the cost is high.Besides,usually farmers do not know the species of agricultural pests in the photographs or videos,which results in the systems cannot find the domain experts quickly and accurately.Crowdsourcing enables task publishers to hire hundreds or thousands of workers quickly and cheaply and to complete the tasks in a short period of time.(2)Recommending solutions of preventing agricultural pests based on crowdsourcing.Farmers accumulate rich experience of preventing agricultural pests as they plant crops all year round.In addition,Seed and pesticide distributors as well as agricultural service personnel also have certain professional knowledge in preventing agricultural pests.They are able to make more effective diagnostic plans after reviewing the photographs or video with short text.This kind of recommendation approach has short cycle and easy to be understood by farmers.(3)Developing a crowdsourcing-based system of preventing agricultural pests.Relying on 4G mobile networks and smart phones,a crowdsourcing-based system of preventing agricultural pests can be developed by combining SQL 2005 and computer languages,such as C#,Java,JQuery.The system changes the role of farmers from passive acceptance to active participation by employing the active participation from the front-line farmers.Therefore,the rich experience from front-line farmers are fully utilized and the prevention efficiency is further improved.