- 學位論文
行車紀錄器之限速路標辨識與應用
An Application of Event Data Recorder to Speed Sign Recognition
摘要
在本論文中，我們提出一個方法利用行車記錄器錄到的畫面，將限速路標偵測出來，再與穿戴式裝置做結合，將辨識出的結果投影於安全帽護目鏡上，來達到提醒的效果，希望能降低因超速而造成的交通事故。 本研究分為四部分。第一部分我們將行車記錄器中的影片截取成影像，做後續影像處理的步驟。第二部分將截取出的影像做前處理，包括運用imsubtract函數突顯紅色、利用Otsu演算法找出最佳閥值、二值化。第三部分利用每個區塊像素個數、長寬比及面積篩選出路標，若路標有歪斜做角度修正，以及運用特徵點辨識路標內的數字。最後一部分利用matlab模擬研究步驟並呈現實驗結果。 在本論文中，主要貢獻為下列幾點： 1.安全性：將限速投影於安全帽護目鏡上，讓騎士不必分心看路旁的路 標，更加安全。 2.簡單性：騎乘機車時行車記錄器接收的畫面非常多，我們運用 imsubtract函數降低運算的複雜度。 3.特徵性：為了能更準確的抓取圓形，我們運用圓形路標特徵點來抓取。 4.趨勢性：因應科技的腳步，我們將結果與穿戴式裝置做結合。
並列摘要
In this thesis, we will propose a method to use the pictures of event data recorder to detect speed limit signs combining with wearable devices. The results of identification will be projected on the helmet visor. It can reach the purpose reminding motorcycle riders to reduce traffic accidents caused by speeding. This study is divided into four parts. The first part, we will intercept event data recorder films into images to make the subsequent image processing steps. In the second part the intercepted images will be used to do image preprocessing, including the using of imsubtract function to highlight red and Otsu method to find the optimal threshold to do binarize. The third part we use the number of pixels, aspect ratio and area to screen road signs in each block. If the speed limit sign has the deflection, we have to do the angle correction, and use feature points within digital recognition. In the last part, simulation procedures are executed by matlab to present the results. In this thesis, we have some contributions as follows: 1. Security：The speed limits projected onto helmet visor to let riders without being distracted by watching speed sign to increase safety. 2. Simplification：Event data recorder receives picture very much and we use imsubtract function to reduce computation complexity. 3. Characterization：In order to more accurately fetch round, we utilize feature points of speed limit sign. 4. Trending：In order to keep up with the pace of technology, we let the proposed algorithms to combine with a wearable device.