A fault-tolerant computing method for Xdraw parallel algorithm

Dou, Wanfeng; Li, Yanan

doi:10.1007/s11227-018-2321-x

Cited by 6 publications

(2 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Dou et al revised the parallel algorithm for XDraw by the analysis of the data dependent relationship between layers. A fine-granularity scheduling strategy is applied to improve the efficiency of the viewshed computation on the process-level and thread-level [20,29,30].…”

Section: Improvements Of Xdrawmentioning

confidence: 99%

HiXDraw: An Improved XDraw Algorithm Free of Chunk Distortion

Zhu

et al. 2019

IJGI

View full text Add to dashboard Cite

Viewshed analysis is of great interest to location optimization, environmental planning, ecology and tourism. There have been plenty of viewshed analysis methods which are generally time-consuming and among these methods, the XDraw algorithm is one of the fastest algorithms and has been widely adopted in various applications. Unfortunately, XDraw suffers from chunk distortion which greatly lowers the accuracy, which limits the application of XDraw to a certain extent. Previous works failed to remove chunk distortion because they are unaware of the underlying contribution relationship. In this paper, we propose HiXDraw—an improved XDraw algorithm free of chunk distortion. We first uncover the causation of chunk distortion from an innovative contributing perspective. Instead of recording LOS (line-of-sight) height, we use a new auxiliary grid to preserve contributing points. By preventing improper terrain data from contributing to determining the visibility, we significantly improve the accuracy of the outcome viewshed. The experimental results reveal that the error rate largely decreases by 65%. Given the same computing time, HiXDraw is more accurate than previous improvements in XDraw. To validate the removal of chunk distortion, we also present a pillar experiment.

show abstract

Section: Improvements Of Xdrawmentioning

confidence: 99%

HiXDraw: An Improved XDraw Algorithm Free of Chunk Distortion

Zhu

et al. 2019

IJGI

View full text Add to dashboard Cite

show abstract

“…Zhu et al proposed the HiXDraw algorithm, using contribution points instead of reference points in XDraw as the basis for computing the visibility of the target points, thus avoiding the interference of irrelevant visible points (referred to as 'chunk distortion' in their work) on computational results [10]. To further improve the efficiency of the algorithms, many parallel computing methods have been proposed and applied to viewshed analysis as well, showing great potential [14]. Apart from these traditional algorithms or their enhanced versions, Tabik et al [15] employ horizon calculation methods [16] in viewshed analysis.…”

Section: Introductionmentioning

confidence: 99%

HiPDERL: An Improved Implementation of the PDERL Viewshed Algorithm and Accuracy Analysis

Cheng¹,

Dou²

2022

IJGI

Self Cite

View full text Add to dashboard Cite

Terrain viewshed analysis based on the digital elevation model (DEM) is of significant application value. A lot of viewshed analysis algorithms have been proposed, including R3 as the accurate one and others as efficient ones. The R3 algorithm is accurate because of its comprehensive but time-consuming computation, while the others are efficient due to proper approximation. However, no algorithm is capable of taking advantage of both until one algorithm is proposed, which is based on a ‘proximity-direction-elevation’ (PDE) coordinate system and named the PDE spatial reference line (PDERL) algorithm. The original research proves the PDERL algorithm is perfectly accurate by theory and experimental results, in comparison with R3 as standard, and even more efficient than R3. However, the original research does not mention the cases where the observer is placed on grid points, and the original implementation does not produce very accurate results in practice. It is important to find out and correct the errors. In this paper, a checking algorithm for PDERL is proposed to allow further investigation of errors. With the fundamental ideas of PDERL unchallenged, an improved implementation of the PDERL algorithm is proposed, named HiPDERL. By experimental results, this paper proves HiPDERL utilizes the potential of PDERL on accuracy at the cost of a little efficiency when the observer is placed on grid points.

show abstract