论文撰写ISM2015

[baidut]

Important Dates

April 17, 2015 Workshop Proposal Submission May 1, 2015 Workshop Notification July 17, 2015 Paper Submission September 4, 2015 Paper Notification September 25, 2015 Camera-Ready Paper Submission

The IEEE International Symposium on Multimedia (ISM) 2015 (http://www.ieee-ism.org/ ) builds on the success of the past ISM conferences in the field of multimedia computing to share information and identify emerging research topics on areas such as Multimedia Systems and Architectures, Multimedia Communications and Streaming, Multimedia Interfaces, Multimedia Coding/Processing/Quality Measurement, Multimedia Security, Multimedia Content Understanding/Modeling/Management/Retrieval, and Multimedia Applications, etc.

System and Architectures Yonggang Wen, Nanyang Technological University, Singapore Ju Shen, Assistant Professor at the University of Dayton, USA

Communications and Streaming Balakrishnan Prabhakaran, University of Texas at Dallas, USA Mea Wang, University of Calgary, Canada

Media Interfaces Matthew Cooper, FX Palo Alto Laboratory, USA Joaquim Jorge, INESC-ID / Técnico / U Lisboa, PORTUGAL

Media Representation, Processing and Quality Measurement Siwei Ma, Peking University Fernando Pereira, Instituto Superior Técnico, Universidade de Lisboa - Instituto de Telecomunicações

Security Geetha Thamilarasu, University of Washington Bothell, USA Pedro Comesaña-Alfaro, University of Vigo, Spain

_Content Understanding, Modeling, Management, and Retrieval _ Qingming Huang, University of Chinese Academy of Sciences, China Song Gao, Google Inc., USA

Mobile media Rongrong Ji, 厦大教授 个人主页 Vijay Chandrasekhar, Researcher, Institute for Infocomm Research, Singapore Huizhong Chen, Ph.D Candidate, Stanford University

Applications 刘烨斌 百度百科 个人主页 Guan-Ming Su, Dolby Labs, USA

题目 信息量大，具体，突出关键：阴影的鲁棒性、基于特征的检测 A novel approach for vision-based (shadowy) road detection Novel Lane Detection using HSV Color Space

标题大小写问题 篇幅：一般至少4也,最多7页,

APA是文科参考文献的格式 MLA是理科参考文献的格式 Unless there are six authors or more give all authors' names; do not use “et al.”.

We suggest that you use a text box to insert a graphic (which is ideally a 300 dpi TIFF or EPS file, with all fonts embedded) because, in an MSW document, this method is somewhat more stable than directly inserting a picture. To have non-visible rules on your frame, use the MSWord “Format” pull-down menu, select Text Box > Colors and Lines to choose No Fill and No Line.

[baidut]

全文结构安排：

• 介绍和行文结构说明
• 车道检测的一般步骤和各种方法，本文在各个步骤上利用HSV颜色空间做了改进
• 图像预处理（新的水平线检测方法划分ROI，进一步地路面区域提取做掩码防止周围干扰）
• 特征提取（采用融合特征进行提取，算法评估）
• 模型拟合（各种方法，选用最简单的）
• 实验结果
• 结论

本文的创新点展开叙述，重点不能偏离，着重介绍针对阴影等特殊情形的处理以证明其鲁棒性

I. INTRODUCTION II. RELATED WORK(Previous work) III. OUR APPROACH IV. (COMPARISONS AND EXPERIMENTS)EXPERIMENTAL RESULTS V. CONCLUSION(and Future/Further Work) REFERENCES

The paper is structured as follows. Section II presents related work. Section III describes the six road marking extraction algorithms and the two variants which are evaluated. The database built for the evaluation, the ground truth labeling process and the evaluation metrics are detailed in Section IV. Section V discusses the exper- imental results obtained after processing the reference database with the six algorithms. Finally, Section VI gives concluding remarks and perspectives.

[baidut]

简介部分删节：

Many excellent works have been done in the field of detecting longitudinal lane-markings using a monocular vision sensor. 各种方法的利弊比较 基于模型的方法，IPM可以有效地防止干扰并解决透视问题但需要摄像头校准信息

Road detection methods can be classified into two groups: feature based and model based methods. The former utilize the visual feature clue, mainly is edge and intensity However, since visual features vary with the illumination, weather condition and distance of the region, feature-based algorithm is restricted to the illuminant variation and laterally adjacent regions. On the other hand, conventional geometric estimation-based approaches, not relying on visual features, are inaccurate and susceptible to pitch or lateral movement. 主要现存问题以及现有处理方法 Cast shadows on road surface are major source of clutter due to the intensity edges they produce. Shadows from nearby trees and buildings may create misleading edges and texture on the road. In some cases, like when the host vehicle comes out of a tunnel there are abrupt changes of several orders of magnitude in the illumination level, leading to over exposed image. 通过方向过滤阴影 In [3] only edges that align with the presumed road boundary directions were extracted, and hence most shadow-related clutter is filtered. linear-parabolic model performs well in the presence of sparse shadows (such as irregular shadows cast by trees), it may present erroneous results if strong aligned shadows appear close to lane markings. In such cases, edges introduced by shadows may be stronger than edges related to lane markings, specially in dashed markings, as illustrated in Fig. 18(b).

传统方法基本步骤叙述，来自ROMA

[baidut]

道路检测部分删节

Our proposed gradient-enhancing conversion method produces a new gray-level image from an RGB color image based on linear discriminant analysis.

车道检测部分删节

The left and right boundary of each row and the lane-marking width can be calculated easily as follows, which is shown in XX, Perspective effect

DLD方法描述

which are a variable threshold and a function of input intensity to robustly extract lane markings under various illumination situations. Then, if the pixel-distance between local maximum and minimum is the estimated lane width that is set in advanced, we regard it as a lane feature Multi-lane detection based on accurate geometric lane estimation in highway scenarios

[baidut]

实验结果 Results of HSILMD including tunnel with bright lighting, an overcast day with both white and yellow lane-markings, and crushed-stone roads with red and yellow markings are shown in Fig. 14. Detected lane-markings on the images of either color are superimposed as with bright green and bright red for ease of observation. Corresponding computation time consumption of each stage of the results in millisecond (ms) is recorded on Table I.

[baidut]

Further work

本文提出的两个方法相对独立，正确的道路边界提取可以有效地辅助车道标记线的识别，同时错误的提取也不会对此造成严重干扰。后续可以探究如何更好地将两者结合起来。如可以通过假设验证的方式提高检测的效率和准确性。基于模型的方法可以添加到现有的方法中。在道路边界检测成功时通过模型信息找到车道标记的可能位置，从而可以在更快地检测车道标记的同时避开阴影等图像干扰，得到更加健壮准确的结果；另一方面，如果道路边界检测错误或失败时，需要能够通过检测到的车道标记线评估错误并进行恢复。 同时，本文所提出的自适应的ROI可以用作视频序列的调整。

Since using contextual information for lane detection will lead to more accurate detection results, future work will integrate the temporal information and geometric transformations of the image to improve the performance of the lane-detection module and accomplish the lane-tracking module. Reference proposed a road boundary detection method based on Hue component. Introduction of geometric transformations of the image, intensity normalization and improvements in object tracking will be shown in future publications.

[baidut]

更换去噪说明图片，没有体现出去除噪声【此段需要改进描述】 图片ROI 改为 2/3 区域 最后的修改：去除基金说明

论文后续工作： 测量运算时间，matlab代码评测工具，简单的tc方式太麻烦 The main problem oflane detection is the contradiction between the algorithm's complexity and the processing speed. The contradiction is enhanced in the complex real world where a lot of water areas, shadows and obstacles are. To solve the problem, 绿色标出检测到的marking部分- 更多更新的文献

[baidut]

论文接收了再通知谭老师

[baidut]

可能需要添加的内容 更多的细节： 左右边界点以及每一行车道线宽度的计算（基于相似关系）Computation of the left and right boundary of each row 量化的评估和对比

论文修改 ISM修改：参考文献顺序 在related work一节依次给出，一般学术文章的参考文献数量以20-40篇为宜。参考文献要与文章直接相关 添加水平线假设，说明图像水平，添加滤波作用说明和改进