中文摘要：

根據臺北市政府警察局《臺北市警務統計年報》顯示，機車竊盜犯罪是近年來竊盜犯罪的主要類型。同時，目前鮮有討論台灣機車竊盜犯罪與街道環境因素關係的量化研究。因此，通過量化探討街道環境因素與機車竊盜犯罪的關係，是一個兼具學術價值與社會意義的研究題目。
本研究通過文獻回顧，藉鑒情景犯罪預防理論及現有城市犯罪行爲研究，整理出一份機車竊盜犯罪環境特徵檢測清單。再根據深度學習圖像語義分割理論，結合現有對谷歌街景圖像數據的研究案例，提出利用街景圖像語義特徵分析機車竊盜犯罪行爲的方法。
此外，本研究蒐集了台北市 180 個日間機車竊盜犯罪記錄點位與 147 個無記錄點位用於統計分析。研究結果表示，在 95% 顯著水準上，日間機車竊盜與犯罪的發生與 13 個環境因素存在相關關係。同時，本研究建立了台北市日間機車竊盜犯罪與 7 個環境因素的二元羅吉斯回歸模型，在測試集上取得 77.27% 的精度。

英文摘要：

According to the Taipei City Police Department’s "Taipei Police Statistics Annual Report", motorcycle theft is the main type of theft in recent years. At the same time, there are few discussions about the quantitative study of the relationship between Taiwan's motorcycle theft and street environmental factors. Therefore, to quantitatively explore the relationship between street environmental factors and motorcycle theft crime is a research topic with both academic value and social significance. Through literature review, this study draws on the theory of situational crime prevention and the existing research on urban crime, and compiled a checklist of environmental characteristics of motorcycle theft crime. According to the theory of deep learning image semantic segmentation, combined with existing research cases on Google street view image data, a method for analyzing the crime of motorcycle theft by using the semantic features of street view images is proposed. In addition, this study collected 180 day-time motorcycle theft crime records and 147 unrecorded points in Taipei City for statistical analysis. The results of the study indicate that at a significant level of 95%, the occurrence of motorcycle theft and crime during the day is related to 13 environmental factors. At the same time, this study established a dual logistic regression model of Taipei City's daytime motorcycle theft crime and 7 environmental factors, and achieved an accuracy of 77.27% on the test set.

文章連結：

https://hdl.handle.net/11296/n32mz2

資料來源：

臺灣博碩士論文知識加值系統