异态异构农业模型服务化封装方法研究

Research on service-oriented encapsulation method for heterogeneous agricultural models

关键词:  农业模型  异态异构模型  模型资源共享  标准化服务封装

Key words:  agricultural models  heterogeneous and diverse models  model resource sharing  standardized service encapsulation

摘要:

【目的】 农业模型涉及多学科与多领域，是智慧农业的核心要素。屏蔽农业模型开发语言、调用运行方式等异态、异构特性，实现农业模型资源高效集成调用与共享，对于农业系统要素科学农业管理和决策，促进农业可持续、高效与安全发展具有重要意义。【方法】 文章设计了模型描述接口、模型执行接口和模型部署接口等标准化封装接口，并在此基础上提出了针对COM组件、Jar包、R语言、Python语言以及DLL等不同形态模型运行体的标准化服务封装方法；以Python形态与DLL形态实现的马铃薯智能化推荐施肥模型为例，验证异态异构模型服务化封装方法的可行性与实用性。【结果】 （1）封装后形成了模型标准化描述文档，完成了模型运行入口函数转换，增加了第三方依赖环境库，最终可通过网络请求直接运行模型，降低了模型使用成本；（2）以云南省为例，对封装后的标准模型服务进行调用运行，输入马铃薯目标产量，可获得该目标产量下最佳施肥量推荐，验证了异态异构模型服务化封装方法的可行性与实用性。【结论】 该方法能够屏蔽模型在语义、数据、形态、运行环境等方面的差异，有效促进农业模型资源在网络环境下复用共享，增强智慧农业领域模型资源管理能力。

Abstract:

［Purpose］ Agricultural models，spanning across various disciplines and domains，constitute the cornerstone of smart agriculture. Shielding the heterogeneous attributes，such as development languages and invocation methods of agricultural models，is crucial for enabling efficient integration and sharing of agricultural model resources. This is paramount for facilitating scientific agricultural management and decision making，thereby fostering sustainable，efficient and secure agricultural development.［Method］ This paper designed standardized encapsulation interfaces，including model description interfaces，model execution interfaces，and model deployment interfaces. Based on this，it proposed standardized service encapsulation methods tailored for different forms of model runtime bodies such as COM components，Jar packages，R language，Python language，and DLLs. The feasibility and practicality of the heterogeneous model service encapsulation methods were validated using an example of a potato intelligent fertilization recommendation model implemented in both Python and DLL forms.［Result］ （1）After encapsulation，the model generated standardized description documents，completed the transformation of model entry point functions，and integrated third-party dependency libraries. As a result，the model could be directly executed via network requests，reducing the cost of model usage. （2）Taking Yunnan Province as an example，the standardized model services post-encapsulation were invoked and operated. Inputting the target potato yield，optimal fertilizer recommendations for that specific yield were obtained，validating the feasibility and practicality of the heterogeneous model service encapsulation methods.［Conclusion］ This method effectively shields discrepancies in semantics，data，form，and runtime environments of models，thereby promoting the efficient sharing and reuse of agricultural model resources in network environments. It enhances the management capability of model resources in the domain of smart agriculture.