面向智能农业的山地果园监测与数据挖掘技术研究及其应用

It is a critical part to do data mining and data analytic on the orchard environment data in order to predict the growing conditions of orchard trees and make corresponding decisions. Our project is mainly about the environment data acquisition, data pre-processing, and applying the deep learning techniques to get the insights of the data. Based on the data analysis results and the real orchard environment, we can provide professional and accurate guidance for orchard management. Data acquisition and analytic is the most important part for agriculture intelligence. For this part, our project first proposed a multi-sinks and hybrid sensors network structure to avoid the problem of disconnected brought by various terrain environment. Second, based on multiple evaluation indicators such as energy efficiency, load balancing, reception rate, and time-to-live of sensor networks, we will improve particle swarm optimization algorithm to optimize network clustering and routing model, and design an efficient mobile assistance data collection strategy. For the data mining part, our project applies deep learning techniques to do analysis on the inner relationships and corrections of the data from multiple sensors. Fully connected network is used for multi-label classification. A new and advanced intelligent orchard environment monitor system is built on the basis of all those techniques. Our project is aproposing a new method to do orchard management to the community and providing both important theoretical and practical values for modern agriculture research.

挖掘果园环境监测数据以预测果树生长状况并制定应对保护措施，是当前果园信息化和智能化的发展需求。本项目立足于山地果园环境，研究环境感知数据采集、预处理及其基于深度学习的分析与预测方法，为果园种植提供科学指导。感知数据采集与挖掘是智能农业中的关键技术。在感知数据采集方面，本项目首先针对果园复杂地形环境，提出多汇聚节点部署和混合传感器网络结构，避免了地形和植被干扰造成连通性问题；其次，基于传感器网络能效、负载均衡、接收率、生存时间等多种评价指标，改进粒子群算法以优化网络分簇和路由模型，并设计高效移动协助数据采集策略。在果园环境数据挖掘方面，本项目拟构建深度学习模型，分析各类型传感器采集数据内在相关性，再设计全连接网络实现多标签分类；在以上研究基础上，开发面向智能农业应用的山地果园环境实时监测系统平台。本项目不仅为果园管理智能化、精准农业等提供新的学术思路，而且对农业工程具有重要的理论和应用价值。

采用信息化手段对果园环境数据监测与果树生长状况的预测，是当前果园大规模建设发展的内在需求；由于山地果园地形与植被环境的特殊性，环境直接干扰了对数据采集的完整性和实时性，感知数据的分析预测也是智慧农业发展的关键问题之一，针对以上问题本项目立足于山地果园环境，研究环境感知数据采集、预处理及其基于深度学习的分析与预测方法，为果园种植提供科学指导。在感知数据采集方面，本项目首先针对果园复杂地形环境，提出多汇聚节点部署和混合传感器网络结构，避免了地形和植被干扰造成连通性问题；其次，基于传感器网络能效、负载均衡、接收率、生存时间等多种评价指标，改进粒子群算法以优化网络分簇和路由模型，并设计高效移动协助数据采集策略。在果园环境数据挖掘方面，本项目构建深度学习模型，分析各类型传感器采集数据内在相关性，再设计全连接网络实现多标签分类；在以上研究基础上，本项目开发了面向智能农业应用的山地果园环境实时监测系统平台。本项目成果不仅为果园管理智能化、精准农业等提供新的学术思路，而且对农业工程具有重要的理论和应用价值。