Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71935
Title: | 有限狀態機模型應用於蜂箱內蜜蜂運動行為分析 Application of Finite State Machine Model for the Analyses of Honeybee Movement Behavior in Beehive |
Authors: | Kuang-Chin Wu 吳匡晉 |
Advisor: | 林達德(Ta-Te Lin) |
Keyword: | 有限狀態機模型,行為模式,深度學習,機器學習,影像處理, Finite State Machine Model,Pattern Behavior,Deep Learning,Machine Learning,Image Processing, |
Publication Year : | 2018 |
Degree: | 碩士 |
Abstract: | 世界上的糧食與經濟作物多依賴蜜蜂的傳遞花粉,但近年所發生的蜜蜂族群崩潰症候群,造成大量的蜜蜂族群無故消失,對於農業生產影響重大,因此極需透過研究蜜蜂行為與建置有效的研究工具以了解該現象之成因。在探索成因之前,必須先對蜜蜂生態有更深的理解,而蜜蜂於蜂巢內透過不同的分工及不同的外力因素會產生不同的行為,這是蜜蜂生態裡相當重要的資訊。基於此,本研究以蜜蜂巢內行為影像監測系統進行實驗,將欲觀察的蜜蜂貼上防水文字標籤,使用標籤辨識與軌跡追蹤等影像處理技術,擷取並紀錄樣本蜜蜂於蜂巢內的運動軌跡(蔡, 2016)。藉由此標籤辨識技術,可將蜜蜂分群標記,透過軌跡追蹤技術,得到不同族群的蜜蜂運動軌跡,建立一個有限狀態機模型對其軌跡進行分析,獲得蜜蜂的行為模式。並利用此行為模式資訊,透過機器學習及深度學習的方法,建立一套蜜蜂行為軌跡分析方法,分析不同分工及不同外力影響下蜜蜂的行為差異。
本研究設計三種實驗進行驗證此分析方法,第一部份為利用有限狀態機模型所分析出來的行為模式資訊,分析內、外勤蜂的行為差異。由實驗結果得知,利用此行為模式資訊可成功分類內、外勤蜂,同時驗證了有限狀態機模型建立的合理性。第二部分為不同箱但同種類的蜜蜂行為模式比較,由實驗結果得知,只要同是內勤蜂或是外勤蜂,即使生活在不同箱的蜜蜂,其行為模式的表現依舊大致相同,驗證了有限狀態機模型分析方法的可行性。第三部分為透過餵食含有不同濃度農藥的糖水,觀察其行為與正常蜜蜂的差異。由實驗結果得知,若蜜蜂受到農藥影響,其行為亦會受到影響,且不同濃度造成的差異性也不同。 Honey bees are important pollinator of our agriculture crops. Most of the food we eat daily is depend on honey bee for pollination. However, massive deaths of honey bees have been reported all around the world. This syndrome named colony collapse disorder or CCD. This phenomenon is being cause a significant impact on the agriculture production. Therefore, it is a need to develop an effective tool to measure the causes of CCD. Honey bee’s trajectories inside the beehive are interesting and important information for understand the individual as well as whole colony behavior. In order to analyze the behavior of individual honey bees, a circular paper tag was glued on the thorax of each honey bee for observation. We developed image processing technique for label recognition and tracking, the movement of honey bees inside the hive is recorded and analyzed. This study labeled honey bees by groups: young bees and forager bees; our imaging algorithm tracked all labeled bees and provided the trajectories of individual honey bees; these trajectories later divided into different stages used to feed our finite state machine (FSM) model, the FSM model was used to analyze the trajectories of honey bee, and get the pattern behavior. Machine learning and deep learning models was trained to classify and recognize different groups of honey bees based on their pattern behavior. To evaluate the performance of the system, multiple set-ups were deployed and several experiments were designed. The experiment involves the comparison of movement inside the hive of in-hive and forager bees. And assessment of pesticide effect on honey bee’s behavior inside the hive. Results of the experiment proved the reliable of the FSM model to classify the movements of different groups of honey bees. We also found out that honey bees of the same group share same behavior pattern. The experiment results also indicate that honey bees reflect different to different concentration of pesticide. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71935 |
DOI: | 10.6342/NTU201803831 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 生物機電工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-107-1.pdf Restricted Access | 3.14 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.