CS2125 Paper Review Form - Winter 2019

Reviewer: Yasaman Rohanifar	

Paper Title: SpringRobot: A Prototype Autonomous Vehicle and Its Algorithms for Lane Detection

Author(s): Qing Li, Nanning Zheng, Hong Cheng

1) Is the paper technically correct?
 [*] Yes
 [ ] Mostly (minor flaws, but mostly solid)
 [ ] No

2) Originality
 [ ] Very good (very novel, trailblazing work)
 [*] Good 
 [ ] Marginal (very incremental)
 [ ] Poor (little or nothing that is new)

3) Technical Depth
 [ ] Very good (comparable to best conference papers)
 [*] Good (comparable to typical conference papers)
 [ ] Marginal depth
 [ ] Little or no depth

4) Impact/Significance
 [ ] Very significant
 [*] Significant
 [ ] Marginal significance.
 [ ] Little or no significance.

5) Presentation
 [ ] Very well written
 [ ] Generally well written
 [ ] Readable
 [*] Needs considerable work
 [ ] Unacceptably bad

6) Overall Rating
 [ ] Strong accept (award quality)
 [ ] Accept (high quality - would argue for acceptance)
 [*] Weak Accept (borderline, but lean towards acceptance)
 [ ] Weak Reject (not sure why this paper was published)


7) Summary of the paper's main contribution and rationale for your recommendation. (1-2 paragraphs)

	This paper briefly introduces SpringRobot, a prototype autonomous vehicle produced by the institute of Artificial Intelligence and Robotics in China, and then discusses the adaptive randomized HT (RHT) method for robost and accurate detection of lane markings without manual initialization or priori information under different road environments in length. The adaptive method introduced by the authors provides high accuracy (especially for noisy images). Some techniques to reduce high computational costs (reducing the size of input image), dealing with cluttered scenes (using a grey-edge map with a small threshold value), providing simplicity (reducing the 3-D Hough space to 2 dimensions), and saving memory storage while decreasing computational cost (randomized sampling and adopting iterative coarse-to-fine accumulation and search strategy) have been proposed throughout the paper. The experiments of the authors, based on their laboratory and images provided by Robotics Institute in Carnegie Mellon University,  showed the validity of their method on lane detection.  

	This paper was published in 2004 and therefore, some of the challenges that the authors faces when writing the paper such as using machine learning to detect vehicles, have been mitigated by now. However, although I am not an expert in this field, I argue for this paper's acceptance as I think it provides a good insight for lane detection methods and makes good use of existing techniques which can be beneficial in future works. The main weak point of this paper is its super concise experiments/evaluation and lack of scientific report on their experiments and results which makes it a borderline accept.


8) List 1-3 strengths of the paper.  (1-2 sentences each,identified as S1, S2, S3.)

S1: Though the paper does not fully meet the promises made in the title and abstract and is mostly about the lane marking detection algorithm rather than the autonomous vehicle prototype, it provides thorough and technical information about the lane detection algorithm. The authors demonstrate good knowledge in the area. 

S2: Equations and flowcharts were well-explained and clear. The explanation about parameters and their method was clear and understandable. 

S3: Leveraging both the advantages of adaptive HT and RHT to provide an algorithm with the benefits of both methods (high parameter resolution, small storage requirements, high speed, etc) is an interesting idea that I think was worth elaborating on and can help develop further improved methods later on. 

9) List 1-3 weaknesses of the paper (1-2 sentences each, identified as W1, W2, W3.)

W1: Many parts of the paper are vague. Where some parts of paper provide basic trivial information about autonomous vehicles and image processing, more explanation is required for some other parts (SpringRobot prototype, some lane detection algorithms discussed, etc). Some of the decisions made by the authors are also not clear in the paper. For example, they claim that they found the value of some parameters (w1, w2) by experiment. The reason why these values cannot be obtained by other methods, and their experiments are not elaborated on. 

W2: The organization and flow of paper is not satisfactory. For example, the introduction is basically a related work section with too much unorganized information about previous work, all in one paragraph. Only a short part of introduction in the end discusses the paper and its contributions.

W3: The expertiments section which also included their evaluation was very short and did not contain enough explanations and show of results. The results of the experiment was not mentioned in a scientific way. No statistics and numbers were published from their experiments. This is a weak point in their paper and makes their paper a more theoretical work rather than actual full implementation.  


The paper itself discusses that the lane-detection depends heavily on how distinguishable marking is from its surroundings and if the marking is distinctive, even a simple detector can do the job, yet they have chosen only color to make the distinction whereas more choices like texture, shape, etc (as they argue themselves) are available to be considered as well to make a better call.