智慧蜂箱監測系統之巢重變化模型與音訊分析

Abstract

本研究探討一套智慧蜂箱監測系統的創建與應用，目的在提升養蜂管理。該系統使用多種感測器收集有關蜂箱重量、溫度、濕度、聲音和蜜蜂進出量的資料。數據通過Wi-Fi傳輸到AWS，使用Amazon Quicksight進行實時監控和歷史資料分析。主成分分析（PCA）和其他機器學習方法被用來將每日重量變化分成六個常見圖形和一個額外其他圖形。比較不同分類結果顯示，SVC模型可以在使用九個特徵時達到0.908的F1-score。這個分類模型運用在該系統在監測和分類蜂箱重量圖形上，可使其成為最佳蜂箱管理的工具。另外在長期音訊頻譜與強度分析中發現，巢內音訊強度取決於蜂群旺盛程度與環境溫度。這個發現補足在蜂群密度上的監測。研究也加入溫度變化、季候條件和蜜蜂進出量對重量圖形的影響，提供有關理想蜂箱位置和管理策略的資訊。研究結果提供一種科學方法與物聯網技術來監測蜂群健康和生產力，提高現代養蜂技術的水平。

This research explores the creation and application of a smart beehive monitoring system aimed at improving beekeeping management. The system uses multiple sensors to collect data on hive weight, temperature, humidity, sound, and bee traffic. The data is transmitted to AWS via Wi-Fi, enabling real-time monitoring and historical data analysis using Amazon QuickSight. Principal component analysis (PCA) and other machine learning methods were employed to categorize daily weight changes into six common groups and an additional "Other" category. Comparison of different classification results showed that the SVC model achieved an F1-score of 0.908 when using nine pattern features. This classification model applied in the system for monitoring and classifying beehive weight patterns can make it an essential tool for optimal hive management. Additionally, long-term analysis of audio spectrum and intensity revealed that the intensity of the hive's internal sound depends on the vigor of the bee colony and the ambient temperature. This discovery complements the monitoring of bee colony density. The study also incorporated the effects of temperature variation, seasonal conditions, and bee traffic on weight patterns, providing insights into ideal hive location and management strategies. The findings offer a scientific approach and IoT technology to monitor hive health and productivity, enhancing the level of modern beekeeping techniques.