台灣農業人口高齡化與跨域人才短缺，是農業長期以來的問題。為了解決此問題，政府積極推動農業與科技跨領域整合，藉由科技輔助農業，將傳統農業轉型為智慧農業，以目前市面上農業科技方案進行分析，固定式天氣站及溫室的環境監控系統，不常用於戶外果樹型果園。為了解決上述科技方案的問題，本文運用模糊神經網路、人工智慧影像辨識、機器人作業系統等技術，利用自動化載具在果園間進行移動，藉由感測系統獲得土壤健康狀況，將農務資訊數位化，即時提供果農果園施肥、噴藥等相關建議，進而減少多餘成本投入，達到優化果樹栽培流程之目的。
此一研究成果可以將果農歷年來的經驗法則科學化，提供合理安全用藥與精準施肥建議，優化整體果樹栽培流程，吸引更多願意嘗試，但缺乏經驗的年輕農夫加入這個產業，從本質上解決農業勞動力老化、生產品質問題，促進產業正向循環成長。

In recent years, agriculture has incorporated various sensors to obtain crop data.Nutrients, conductivity, and pH in soil are the key factors affecting the growth of crops.Before using fertilizers, farmers need to master the soil information of the orchard to decide the ratio of nitrogen, phosphorus, and potassium to be fertilized and the amount of water to be irrigated.
In this paper, we adopt an automatic self-propelled vehicle soil identification system to perform path planning in orchards. After arriving at the measurement location designated by the farmer, use the soil probe measurement system to measure soil nutrients in a particular area, data analysis through fuzzy rules, and a back-end knowledge base system.
The system provides farmers with suggestions on fertilization and medication and grasps the regional growth status of the orchard.

Chinese Abstract........i
English Abstract........ii
Acknowledgements........iii
Table of Contents........iv
List of Tables........vi
List of Figures........vii
Chapter 1 INTRODUCTION........1
1.1 Research motivation and purpose........1
1.2 Purpose and Design Scheme........2
1.3 Research Review........2
1.3.1 Simultaneous Localization and Mapping (SLAM)........2
1.3.2 Path planning........3
1.3.3 Adaptive Network-Based Fuzzy Inference System........4
1.3.4 DeepSORT........4
1.4 Description of thesis structure........4
1.5 Demand Trends of Global Agricultural Robots........5
1.6 Literature Analysis and Discussion........6
1.7 Image Recognition Analysis........6
1.7.1 Introduction to YOLOv4........7
1.8 Literature analysis........7
Chapter 2 Hardware Design and Architecture........8
2.1 Overview of hardware design........8
2.2 Mechanical structure design........8
2.3 Power Mechanism Design........16
2.4 Laser pest control system........16
2.4.1 Laser System Hardware........18
2.4.2 Laser Pest Image Recognition System........20
2.5 Soil Nutrient Detection System........21
2.6 Citrus Yield Analysis System........22
2.7 Overview of hardware system architecture........23
2.8 Core of Agricultural Unmanned Vehicle System........24
2.9 Image system unit........25
2.9.1 Environmental Detection System........25
2.10 Chassis drive unit........26
2.11 Power Systems........27
Chapter 3 Software System Architecture........28
3.1 Deep Learning Object Recognition........28
3.1.1 YOLOv4........28
3.2 LeGO-LOAM Simultaneous Localization and Mapping........32
3.3 Navigation and Obstacle Avoidance........33
Chapter 4 Experiments........35
4.1 AI image recognition training pre-work........35
4.1.1 Display card information........35
4.1.2 Opencv and CUDA and CUDNN version models........36
4.1.3 Computer specification information for deep learning training........36
4.1.4 Parameter setting before image recognition training........37
4.1.5 YOLO training process........37
4.1.6 Prepare dataset samples........38
4.1.7 Labeling samples........40
4.1.8 Split the data........41
4.1.9 Change the parameter file in the cfg file........41
4.2 YOLOv4 training process........43
4.2.1 Citrus Dataset Training........43
4.2.2 Mosaic and CutMix data enhancement method in YOLOv4........46
4.2.3 Snail dataset training........50
4.3 Robotic arm control of laser pest control system........54
4.3.1 Principle of laser pest control system........55
4.4 Process of Soil Nutrition Analysis System........57
Chapter 5 Conclusion and Future Outlook........64
5.1 Conclusion........64
5.2 Future Outlook........65
References........66
Extended Abstract

1.Bochkovskiy, A., C.-Y. Wang, and H.-Y.M.J.a.p.a. Liao, Yolov4: Optimal speed and accuracy of object detection. 2020.
2.Veeramani, B., J.W. Raymond, and P. Chanda, DeepSort: deep convolutional networks for sorting haploid maize seeds. BMC bioinformatics, 2018. 19(9): p. 1-9.
3.Dwijotomo, A., et al., Cartographer slam method for optimization with an adaptive multi-distance scan scheduler. Applied Sciences, 2020. 10(1): p. 347.
4.Shan, T. and B. Englot. Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain. in 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 2018. IEEE.
5.Hart, P.E., N.J. Nilsson, and B. Raphael, A formal basis for the heuristic determination of minimum cost paths. IEEE transactions on Systems Science and Cybernetics, 1968. 4(2): p. 100-107.
6.Koenig, S. and M. Likhachev, Fast replanning for navigation in unknown terrain. IEEE Transactions on Robotics, 2005. 21(3): p. 354-363.
7.Fox, D., W. Burgard, and S. Thrun, The dynamic window approach to collision avoidance. IEEE Robotics & Automation Magazine, 1997. 4(1): p. 23-33.
8.Jang, J.-S.J.I.t.o.s., man, and cybernetics, ANFIS: adaptive-network-based fuzzy inference system. 1993. 23(3): p. 665-685.
9.Berners-Lee, M., et al., Current global food production is sufficient to meet human nutritional needs in 2050 provided there is radical societal adaptation. Elementa: Science of the Anthropocene, 2018. 6.
10.R Shamshiri, R., et al., Research and development in agricultural robotics: A perspective of digital farming. 2018.
11.Knoll, F., T. Holtorf, and S. Hussmann. Vegetation index determination method based on color room processing for weed control applications in organic farming. in 2016 IEEE International Instrumentation and Measurement Technology Conference Proceedings. 2016. IEEE.
12.Hussmann, S., et al. Development and evaluation of a low-cost delta robot system for weed control applications in organic farming. in 2019 IEEE International Instrumentation and Measurement Technology Conference (I2MTC). 2019. IEEE.
13.Bochkovskiy, A., C.-Y. Wang, and H.-Y.M. Liao, Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934, 2020.
14.Anderson, N.P., et al., Evaluating soil nutrients and pH by depth in situations of limited or no tillage in western Oregon. 2010.
15.He, K., et al., Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE transactions on pattern analysis and machine intelligence, 2015. 37(9): p. 1904-1916.
16.Liu, S., et al. Path aggregation network for instance segmentation. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
17.Zhang, Z., et al., Bag of freebies for training object detection neural networks. arXiv preprint arXiv:1902.04103, 2019.
18.Yun, S., et al. Cutmix: Regularization strategy to train strong classifiers with localizable features. in Proceedings of the IEEE/CVF international conference on computer vision. 2019.
19.Ghiasi, G., T.-Y. Lin, and Q.V. Le, Dropblock: A regularization method for convolutional networks. Advances in neural information processing systems, 2018. 31.
20.Müller, R., S. Kornblith, and G.E. Hinton, When does label smoothing help? Advances in neural information processing systems, 2019. 32.
21.Marder-Eppstein, E. Navigation Package. Available from: http://wiki.ros.org/navigation.
22.Dellaert, F., et al. Monte carlo localization for mobile robots. in Proceedings 1999 IEEE international conference on robotics and automation (Cat. No. 99CH36288C). 1999. IEEE.
23.Thrun, S., Probabilistic robotics. Communications of the ACM, 2002. 45(3): p. 52-57.