本論文採用谷歌圖像辨識技術設計與實現兩個車內人與寵物遺留通知系統。系統一核心組件為谷歌AIY視覺套件及LINE Notify服務；谷歌AIY視覺套件由Raspberry Pi Zero WH開發板、Vision Bonnet及Raspberry Pi Camera V2組成。系統二核心組件為FLIR Lepton Dev Kit熱影像開發模組、Raspberry Pi 3B+開發板、谷歌雲端視覺 API及Line Notify服務。LINE Notify為LINE通訊軟體的通知功能，可以讓我們即時的傳送警告訊息至使用者手機。
    系統一展示了使用內建的谷歌視覺人、狗、貓辨識模組之谷歌AIY視覺套件，由樹莓派相機V2版拍下的照片進行人、狗、貓辨識，若辨識出人、狗、貓類別資訊，則即時將該訊息及圖片傳遞至家人LINE通訊軟體群組內，以便家人及時獲取資訊和處理，避免人、狗、貓車內熱中暑意外事件的發生。
    系統二展示了由FLIR Lepton Dev Kit熱影像開發模組拍下熱影像並且成功達成人物及寵物偵測；使用谷歌雲端視覺API，辨識相片內物件類別，回傳至樹莓派3B+開發板判斷特徵清單內有無人、狗、貓相關關鍵字；若有，則回傳之特徵文字訊息末段加上警告訊息，並即時將該訊息及圖片傳遞至家人LINE通訊軟體群組內通報。此外，針對有嬰兒安全座椅的情況下，探討系統在車內最合適的安裝位置。

This thesis presents the design and implementation of two notification systems for human and pets left inside a car, which employs Google's image recognition technologies. The core components of the first system have a Google AIY Vision Kit, and the LINE Notify service, where the Google AIY Vision Kit consists of a Raspberry Pi Zero WH development board, a Vision Bonnet, and a Raspberry Pi Camera V2. The second system includes the core components of a FLIR Lepton Dev Kit, a Raspberry Pi 3B+ development board, and the Google Cloud Vision API and Line Notify services. The LINE Notify service is a notification function for the LINE that allows us to send alert messages to users' mobile phones instantly.
The first system demonstrated the Google AIY Vision Kit using the built-in classification module, which can identify whether there’s a dog, cat, or person in an image. The photos taken by Raspberry Pi Camera V2 are used to identify people, dogs, and cats. If the system detects them, the alert message and the picture will be immediately transmitted to the family LINE group, so that the family can obtain information and processing in time to avoid the occurrence of car heatstroke accidents in people, dogs, and cats.
The second system demonstrated the thermal image taken by a FLIR Lepton Dev Kit and successfully achieved the person and cat detection, which utilizes the Google Cloud Vision API to identify the object features in the photo and return them to the Raspberry Pi 3B+ development board to judge the feature list with people, dog, cat related keywords. If there is any, the end of the return message will be added with a warning message, and the return message and the picture are immediately transmitted to the family LINE group for notification. Moreover, in the case of a baby safety seat, we explored the most suitable installation position of the system in a car.

目錄
誌謝 I
中文摘要 II
ABSTRACT III
目錄 IV
圖目錄 VI
第一章 緒論 1
1.1 前言 1
1.2 研究動機 2
1.3 論文架構 2
第二章 車內熱中暑預防與先前技術介紹 3
2.1 車內熱中暑 3
2.2 現行市場上相關產品 3
第三章 關鍵元件及使用平台介紹 8
3.1 谷歌視覺辨識模組 8
3.1.1 Raspberry Pi Zero WH 9
3.1.2 Vision Bonnet 10
3.1.3 Raspberry Pi Camera V2 10
3.2 LINE及LINE Developers 11
3.3 Google Cloud Platform及Cloud Vision API 12
3.4 FLIR Lepton Dev Kit 紅外線熱成像開發套件 14
3.5 Raspberry Pi 3B+ 16
第四章 系統架構及實驗 18
4.1 系統架構 18
4.2 實驗設置與結果 20
4.2.1 系統一實驗	20
4.2.2 系統二實驗 27
第五章 結論與未來展望 40
5.1 結論	40
5.2 未來展望	41
參考文獻 42

圖2.1 Evenflo SensorSafeTM技術 4
圖2.2 通用汽車儀表板上的提醒警告顯示 5
圖2.3 駕駛員的小幫手感測器的配件 5
圖2.4 兒童車內熱中暑預防系統方案 6
圖3.1 谷歌視覺辨識模組 8
圖3.2 Raspberry Pi Zero WH開發板 9
圖3.3 Vision Bonnet 10
圖3.4 Raspberry Pi Camera V2 10
圖3.5 LINE 11
圖3.6 LINE Developers 網頁 12
圖3.7 Google Cloud Platform首頁 12
圖3.8 Cloud Vision API 13
圖3.9 FLIR Lepton Dev Kit 紅外線熱成像開發套件 14
圖3.10 FLIR Lepton Dev Kit拍攝之相片 14
圖3.11 FLIR TG165熱顯像儀拍攝之相片 15
圖3.12 FLIR TG165熱顯像儀 15
圖3.13 Raspberry Pi 3B+開發板 16
圖4.1 系統一架構 18
圖4.2 系統二架構 19
圖4.3 系統一操作流程圖 20
圖4.4 辨識前照片(中控台)	21
圖4.5 辨識後照片(中控台)	22
圖4.6 LINE Notify接收訊息及圖片的畫面	23
圖4.7 貓的辨識結果 23
圖4.8 狗的辨識結果 24
圖4.9 辨識後照片(右A柱) 25
圖4.10 辨識後照片(右B柱) 25
圖4.11 辨識後照片(右C柱) 26
圖4.12 系統二操作流程圖	27
圖4.13 LINE Notify接收訊息及圖顯示偵測到人 29
圖4.14 LINE Notify接收訊息及圖顯示無偵測到人或寵物 30
圖4.15 LINE Notify接收訊息及圖顯示偵測到貓31
圖4.16車輛ABC柱位置 31
圖4.17 車內熱影像(人員位於前座右側)及辨識到的特徵 32
圖4.18 車內熱影像(人員位於前座左側)及辨識到的特徵 33
圖4.19 車內熱影像(人員位於後座中間)及辨識到的特徵 33
圖4.20 車內熱影像(人員位於後座右側)及辨識到的特徵 34
圖4.21 車內熱影像(人員位於後座左側)及辨識到的特徵 34
圖4.22 車內熱影像(人員位於前座右側)及辨識到的特徵 35
圖4.23 車內熱影像(人員位於後座右側)及辨識到的特徵 35
圖4.24 車內熱影像(人員位於後座左側)及辨識到的特徵 35
圖4.25 車內熱影像(人員位於前座左側)及辨識到的特徵 36
圖4.26 車內熱影像(人員位於後座中間)及辨識到的特徵 36
圖4.27 車內熱影像(人員位於前座右側)及辨識到的特徵 37
圖4.28 車內熱影像(人員位於後座右側)及辨識到的特徵 37
圖4.29 車內熱影像(人員位於前座左側)及辨識到的特徵 37
圖4.30 車內熱影像(人員位於後座左側)及辨識到的特徵 38
圖4.31 車內熱影像(人員位於後座中間)及辨識到的特徵 38

[1] Null, J. (2019/04/23). Heatstroke Deaths of Children in Vehicles. Retrieved from NoHeatstroke.org, website: http://www.noheatstroke. org/
[2] American Academy of Pediatrics. (2018/07/18). Prevent Child Deaths in Hot Cars. Retrieved from Healthy Children.org, website https://www.healthychildren.org/English/safety-prevention/on-the-go/Pages/Prevent-Child-Deaths-in-Hot-Cars.aspx
[3] McLaren, C., Null, J., and Quinn, J. (2005, July), “Heat Stress From Enclosed Vehicles: Moderate Ambient Temperatures Cause Significant Temperature Rise in Enclosed Vehicles. American Academy of Pediatrics,” 116(1), 1-6.
[4] National District Attorneys Association. Hyperthermia Statutes: Children Unattended In Vehicles (updated August 2014). National District Attorneys Association.org, website: https://ndaa.org/ programs/child-abuse/state-statutes/
[5] 國軍花蓮總醫院, “中暑衛教,” website: https://805.mnd.gov.tw/ 805Hemodialysis/web/heatstroke01.html
[6] How 4 technologies designed to prevent hot car deaths work,
website: http://abcnews.go.com/US/technologies-designed-prevent
-hot-car-deaths-work/story?id=47991074
[7] Evenflo SensorSafeTM Technology, website: http://www.evenflo.
com/sensorsafe/sensorsafe.html
[8] Emily A. Thomas, “GM's Child-Reminder Feature Is a Simple,
Clever Way to Save Lives,” website: https://www.consumerreports
.org/car-safety/gm-child-reminder-feature-simple-clever-way-to-sav
e-lives/
[9] Driver’s Little Helper, website: https://driverslittlehelper.com/
[10]Muhammad Luqman Bin Abidin, Muhammad Syafiq Bin Zainal,
Nor Azlia Binti Azizan, and Nurul Hidayah Binti Samit, “Child
Vehicular Heat Stroke Prevention System by Controlling Vehicle
Features Using Wireless Technology,” website: https://ir.unikl.edu.my/jspui/bitstream/123456789/14901/1/Child%20Vehicular%20Heat%20Stroke%20Prevention%20System.pdf
[11]Elaine Tianjiao Yang and Andrew Tianxing Yang, “Child Car
Heatstroke Alarm and Prevention System,” US Patent Application
Publication, Pub. No. US 2016/0351037, Dec. 1, 2016.
[12]Google AIY Vision Kit, website: https://aiyprojects.withgoogle.com/vision/
[13]維基百科, “LINE,” website: https://zh.wikipedia.org/wiki/LINE
[14]Google Cloud, “Cloud Vision,” website: https://cloud.google.com/vision/
[15]SparkFun, “FLIR Lepton Hookup Guide,” website: https://learn.
sparkfun.com/tutorials/flir-lepton-hookup-guide
[16]Pexels, website: https://www.pexels.com/
[17]老侯解車, “汽車上的A 柱、B 柱、C 柱是指什麼？他們都有什麼
作用？,” website: https://club.autohome.com.cn/bbs/thread/
21746fc4637341ed/81191919-1.html
[18]M. J. Kallan, D. Durbin, and K. B. Arbogast, “Seating Patterns and
Corresponding Risk of Injury Among 0- to 3-Year-Old Children in
Child Safety Seats,” In PEDIATRIC, May 2008.
[19]道路交通管理處罰條例第三十一條第三項, “小型車附載幼童安
全乘坐實施及宣導辦法, ” March 31, 2017.
[20]J. P. S. Barrera, G. M. Sandoval, G. C. Ortiz, R.N. González, and E.
R. Aguilar, “A Multi-agent System to Avoid Heatstroke in Young
Children Left in Baby Car Seats inside Vehicles,” IEEE International
Conference on Computational Science and Computational
Intelligence, Mar. 2014.
[21]S. Deepthi and P. S. Arun, “Recognition of New Born Babies Using
Multi Class SVM,” IEEE International Conference on Circuit,
Power and Computing Technologies, Apr. 2017.