- 學位論文
基於影像處理做平交道障礙物偵測輔助系統之研究
A Study on the Lending Support of Block Detecting Based on Image Processing Applied to the Highway Crossing System
摘要
近年來隨著交通問題日益嚴重，智慧型運輸系統(Intelligent Transportation System, ITS)的建置已經變成一個關鍵的議題。未來能將平交道資訊以串列(RS485)或者乙太網路的方式傳送給控制中心做“障礙物”分析。系統能立即採取動作或發出警告給火車司機員和鐵路警察採取安全措施以降低損害。我的研究重點是在平交道上的行人，機車和汽車的影像處理系統。 火車行車事故中平交道事故最多，如79年的路竹平交道事故、81年八股頭平交道事故。因車輛熄火或滯留於平交道上，列車來不及煞車造成嚴重事故，為防範此種重大事故，本系統能分析並發出警告，提早通知火車司機員準備煞車。本研究則是透過實際平交道障礙物影像分析，期能有效地將平交道之事故大幅降低。 本論文所提方法主要分為三個階段：障礙物的偵測、擷取及分類。這三階段的設計原理分別根基於色彩資訊、區塊資訊和幾何特性。在偵測階段，首先將色彩空間RGB轉為HSI，以偵測出障礙物的特定區塊，接著利用型態學的方法補強區塊的缺損及去除雜訊。在擷取階段，主要是利用區塊標記技巧偵測出障礙物可能的位置，接著使用邊緣偵測法將障礙物邊緣特徵擷取出來以利後續分類階段利用障礙物幾何外型特徵達成分類目的，亦即辨識出障礙物為汽車、機車或行人。 本系統不僅能增加平交道安全防護也提高了在平交道上的汽車、機車和行人的安全性。最後經由眾多實驗結果證明本論文所提出的方法確實有效且可行。
並列摘要
Due to the heavy traffic in recently year, the image processing system has become a key technique for intelligent transportation system (ITS). Those data information on the highway crossing system can use the serial (RS485) or Ethernet as a transmit media to send the “Block” image to control center to analyze in the future. System can take the immediate action to warning the train driver or railway policeman to take safety procedure to lower down the damage. My research paper focuses on the image processing system mainly on the pedestrian, motorcycles and cars on the highway crossing. The highway crossing usually is the most dangerous place in the railway system, such as accident happened in 1980 at the lu-chu highway crossing and in 1992 at pa-gu-tou highway crossing. When cars powered down and stay still on the highway crossing area, the train driver can not stop the train in time when he saw the staying still car. To prevent such serious accident, the highway crossing block detecting system (HCBD) can analyze and warn the train driver to take actions to speed down or stop the train when it driver near the highway crossing. The research focuses on the image analysis of the highway crossing block in order to reduce the accident of the highway crossing. My research consists of three phases: crossing block detection, extraction and classification. The three phases are based on color information, region data and geometric and shape of object. In the detection phase, the RGB color space is first convert into HSI color space to detect those regions with specific colors of blocks. The morphology technology is then used to fix hole and noise of target. In the extraction phase, region labeling is involved to label candidate regions of blocks. Edge detection is then employed to shape region border. In the classification phase, target can classify by the geometric and its shape. The HCBD system can classify whether the target block is a car, motorcycle or pedestrian after the phases processing. The proposed system can increase traffic safety not only for train, but also for all cars, motorcycles and pedestrians on the highway crossing. Finally, various experiments have been conducted to demonstrate the effectiveness and practicability of proposed method.