WEKO3
アイテム
消失点発見とテクスチャオリエンテーション算出による進路エリア検出
https://doi.org/10.18997/00003750
https://doi.org/10.18997/0000375082f7478d-783f-4922-86a6-b212d9b37cab
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
D-172_jou_k_277.pdf (4.8 MB)
|
|
| アイテムタイプ | 学位論文 = Thesis or Dissertation(1) | |||||||
|---|---|---|---|---|---|---|---|---|
| 公開日 | 2013-10-31 | |||||||
| 資源タイプ | ||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_db06 | |||||||
| 資源タイプ | doctoral thesis | |||||||
| タイトル | ||||||||
| タイトル | Road Area Extraction Based on Vanishing Point Detection and Texture Orientation Estimation | |||||||
| 言語 | en | |||||||
| タイトル | ||||||||
| タイトル | 消失点発見とテクスチャオリエンテーション算出による進路エリア検出 | |||||||
| 言語 | ja | |||||||
| 言語 | ||||||||
| 言語 | eng | |||||||
| 著者 |
Bui, Trung Hieu
× Bui, Trung Hieu
|
|||||||
| 抄録 | ||||||||
| 内容記述タイプ | Abstract | |||||||
| 内容記述 | Road area extraction plays an important role in different areas of computer vision such as autonomous driving, car collision warning and pedestrian crossing detection. Detecting road areas from a single road image is a challenging problem as the detection algorithm must be able to deal with continuously changing backgrounds, different environment (urban, high ways, off-road), different road types (shape, color), and different imaging conditions (varying illumination, different viewpoints and changing weather conditions). Over the past few decades, numerous road area extraction methods have been widely published for urban and highway roads, structured roads, and unstructured roads. In all of these studies, estimating a location of vanishing point (VP) is a key requirement. A set of parallel lines in the world space by perspective projection converges to a common point in image space known as the VP. If the VP can be located correctly, then it is more likely to detect the road area properly. State-of-the-art VP-based road detection methods can be mainly grouped into three categories: edge-based methods, prior-based methods, and texture-based methods. Most edgebased methods take advantage of computational efficiency, and they can be applied to real time systems. However, the disadvantage is that they only work well for structured roads with clear painted lines or distinct road boundaries, while they usually fail in unstructured roads lacking sharply defined, smoothly curving edge. In order to overcome the limitation of these edge-based methods, prior-based methods and texture-based methods for road detection have been widely proposed. Most prior-based methods are robust to varying imaging conditions, road types and sur-rounding environments. However, they often require a large-scale image or video training database and also manual works for labeling VPs for the training stage. Hence, such prior-based methods are inapplicable to detect the road region from a single road image. In contrast, texture-based methods are very effective at detecting road areas for both structured roads and unstructured roads. These methods first search for local oriented textures and then make them vote for the location of the roads VP. In order to segment the road area, a VP-constrained group of dominant edges are detected, and two most dominant edges are selected as the road borders. In general, the disadvantages of these texture-based methods are: i) the computational cost of VP detection process is high, and ii) an estimation error of VP detection which may affect the performance of road area extraction is obtained in some images. In this thesis, our main objectives are: i) to reduce the computational cost and improve the performance of VP detection algorithm, and ii) to propose a new VP-based road area extraction method from a single road image. For this purpose, this thesis has been divided into four chapters. In Chapter 1, the background and some related works are introduced. Chapter 2 is in accordance with the first objective mentioned above. In this chapter, our novel texture-based local soft voting method for VP detection is explained. Firstly, Gabor filters and confidence level function are introduced. Gabor filters are used to calculate the texture orientation at every pixel of the road image, and the confidence level function is used to determine the remaining voters which are useful for the VP voting process by checking the reliability of the obtained texture orientations. After that, a novel local soft voting method is proposed, in which the number of scanning pixels is much reduced to reduce the computational cost, and a new VP candidate region is introduced to improve the estimation accuracy. The proposed VP detection method has been implemented and tested on 1000 road images which contain large variations in color, texture, lighting condition and surrounding environment. The experimental results demonstrate that this new voting method is both efficient and effective in detecting the VP from a single road image and requires much less computational time when compared to a previous voting method. Chapter 3 is in accordance with the second objective stated above. In this chapter, our new VP-constrained road area extraction method is described. The goal of this method is to detect the most immediate straight road part in the direction of the optical axis of the forwarding looking cameras based on estimating two lines (road boundaries) going from the VP and below the VP in the road image. In this method, in order to achieve a one degree level of precision for road boundaries detection, a histogram of 180 angles corresponding to angles of 180 lines going from the detected VP is used. When generating the histogram, the color information of the road image is combined to improve the estimation performance. The proposed road area extraction method has been implemented and tested on 1000 road images which are same as ones used in Chapter 2. The experimental results show that our proposed method performs well in challenging conditions. Finally, Chapter 4 presents our conclusions and future work. | |||||||
| 目次 | ||||||||
| 内容記述タイプ | TableOfContents | |||||||
| 内容記述 | 1 Introduction||2 Vanishing point detection||3 Road area extraction||4 Conclusions | |||||||
| 備考 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 九州工業大学博士学位論文 学位記番号:情工博甲第277号 学位授与年月日:平成25年3月25日 | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Road Area Extraction | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Vanishing Point | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Texture Orientation | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Gabor Fieters | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Confidence level | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Soft voting | |||||||
| アドバイザー | ||||||||
| 延山, 英沢 | ||||||||
| 学位授与番号 | ||||||||
| 学位授与番号 | 甲第277号 | |||||||
| 学位名 | ||||||||
| 学位名 | 博士（情報工学） | |||||||
| 学位授与年月日 | ||||||||
| 学位授与年月日 | 2013-03-25 | |||||||
| 学位授与機関 | ||||||||
| 学位授与機関識別子Scheme | kakenhi | |||||||
| 学位授与機関識別子 | 17104 | |||||||
| 学位授与機関名 | 九州工業大学 | |||||||
| 学位授与年度 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 平成24年度 | |||||||
| 出版タイプ | ||||||||
| 出版タイプ | VoR | |||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||
| アクセス権 | ||||||||
| アクセス権 | open access | |||||||
| アクセス権URI | http://purl.org/coar/access_right/c_abf2 | |||||||
| ID登録 | ||||||||
| ID登録 | 10.18997/00003750 | |||||||
| ID登録タイプ | JaLC | |||||||
