應用幾何形狀的頻率特徵於交通標誌偵測之研究

The Study of Traffic Signals Detection Based on the Frequency Features of Geometric Shapes

江駿逵(Chu-Ta Chiang)

指導教授 ： 黃健興

義守大學/電機資訊學院/資訊工程學系/碩士(2016年)

摘要

交通標誌偵測(Traffic Signals Detection)是偵測交通標誌醒目的顏色與形狀，用以提醒駕駛人前方的路況與道路相關的資訊，其形狀大致分為圓形、三角形、矩形與菱形等四大類，顏色大致有紅色、藍色、黃色為交通標誌的主要顏色，故此研究針對交通標誌進行顏色及形狀的頻率特徵進行分類偵測。
首先利用影像色彩分離出可能的交通標誌區域，並使用影像形態學
(Morphology)來減少影像中的雜訊，接著使用影像切割(Image Segmentation)來找出交通標誌可能的群聚位置，最後利用形狀偵測(Shape Detection)在幾何形狀的頻率域空間上表現特徵，區分交通標誌可能的形狀，綜合以上資訊，根據「醒目色彩的幾何形狀」來找出交通標誌所在位置。
本研究中，定義了各種交通標誌的顏色範圍，各種幾何形狀的頻率特徵，並通過以上資訊來檢測交通標誌。該技術對降低交通事故率，來提高交通安全在智慧運輸系統(intelligent transportation system)上有重要的意義。實驗結果證明，所提出可以實現的方法總體正確率可以達到 86.5%。

關鍵字

形態學 ； 影像切割 ； 形狀偵測 ； 智慧運輸系統 ； 交通標誌偵測

並列摘要

The technique of traffic sign detection, which used to remind drivers the road conditions and relative information is to find the traffic sign from the images according to the color or shape features. The shapes of traffic sign would be divided into four categories: circle, triangle, rectangle and diamond, and the main color of traffic sign are red, blue and yellow. Therefore, we proposed a method of traffic sign detection by color classification and frequency characteristics of geometric shape.
At beginning, we find the candidate traffic sign areas by image color classification, and adopt the technique of morphology to enhance the geometric structure information. Then, the region of interest (ROI) are segmented as the position of traffic signs in images. Then, Fast Fourier Transform (FFT) are used to extract the geometric shape features in frequency domain. Eventually, the traffic signs are defined as “geometric shape with remarkable color” and are located their positions.
In this thesis, we define the color range of each kinds of traffic sign, the frequency feature of each kinds of geometric shapes and detect the traffic sign by above information. This technique is important to intelligent transportation system (ITS) to reduce the traffic accident rate and improve traffic safety. The experimental results demonstrate that the proposed method of implementation can reach an accuracy of 86.5%.

並列關鍵字

Morphology ； Image Segmentation ； Shape Detection ； Intelligent Transportation System ； Traffic Signals Detection

參考文獻

[1]R. Achanta, “SLIC superpixels compared to state-of-the-art super-pixel methods, ”IEEE Trans. Pattern Anal. Mach. Intell, vol. 34, no.11, pp. 2274–2282, Nov. 2012.

[2]V. Gopalakrishnan, Y. Hu, and D. Rajan, “Random walks on graphs for salient object detection in images, ” IEEE Trans. Image Process, vol. 19, no.12, pp. 3232–3242, Dec. 2010.

[6]Virupakshappa. K. , Han.Y., Oruklu. E. “Traffic sign recognition based on prevailing bag of visual words representation on feature descriptors ”IEEE International Conference on Electro/Information Technology (EIT),2015.

[9]Fleyeh.H ,“Traffic sign recognition without color information,” Color and Visual Computing Symposium (CVCS), 2015.

[10]Li Ce, Song Wenyang, Xiao Limei, Hu Yaling, Pan Xing “Salient traffic sign video detection based on hypercomplex frequency domain” 33rd Chinese Control Conference (CCC), 2014.