Point processes for unsupervised line network extraction in remote sensing

Lacoste, Caroline; Descombes, Xavier; Zerubia, Josiane

doi:10.1109/tpami.2005.206

Cited by 142 publications

(180 citation statements)

References 25 publications

Supporting

Mentioning

178

Contrasting

Order By: Relevance

“…Thus, a new configuration s = s\{s} is proposed by the death kernel. We compute the acceptance ratio of the birth kernel α B and the death kernel α D in the same way as proposed in [18], given by…”

Section: Monte Carlo Sampler With Delayed Rejectionmentioning

confidence: 99%

“…Thus, it is an important task to detect lines in many computer vision applications. For example, road network extraction algorithms [18,30] have been developed for remote sensing. To find defects of the road pavement, an adaptive filtering and image segmentation algorithm has been proposed in [5].…”

Section: Introductionmentioning

confidence: 99%

“…Sampling techniques with geometric priors have been exploited to detect multiple line segments in a scene [2,16,18,22,24,28]. The marked point process (MPP) framework [6,7,20,26] is helpful to enforce high level constraints on shape prior.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Marked Point Process Model for Curvilinear Structures Extraction

Jeong¹,

Tarabalka²,

Zerubia³

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper, we propose a new marked point process (MPP) model and the associated optimization technique to extract curvilinear structures. Given an image, we compute the intensity variance and rotated gradient magnitude along the line segment. We constrain high level shape priors of the line segments to obtain smoothly connected line configuration. The optimization technique consists of two steps to reduce the significance of the parameter selection in our MPP model. We employ Monte Carlo sampler with delayed rejection to collect line hypotheses over different parameter spaces. Then, we maximize the consensus among line detection results to reconstruct the most plausible curvilinear structures without parameter estimation process. Experimental results show that the algorithm effectively localizes curvilinear structures on a wide range of datasets.

show abstract

Section: Monte Carlo Sampler With Delayed Rejectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Marked Point Process Model for Curvilinear Structures Extraction

Jeong¹,

Tarabalka²,

Zerubia³

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The precision of the estimator is related to the precision of h. Another possible estimator can be constructed using h + r c instead of h. The same construction can be used even if different cylinders are used Lacoste et al (2005); Stoica et al (2002Stoica et al ( , 2004. As for the sufficient statistics, the distribution of the length of the network may be derived using Monte Carlo techniques.…”

Section: Observed Filamentsmentioning

confidence: 99%

Filaments in observed and mock galaxy catalogues

Stoica

Martínez

Saar

2010

A&A

View full text Add to dashboard Cite

Context. The main feature of the spatial large-scale galaxy distribution is an intricate network of galaxy filaments. Although many attempts have been made to quantify this network, there is no unique and satisfactory recipe for that yet. Aims. The present paper compares the filaments in the real data and in the numerical models, to see if our best models reproduce statistically the filamentary network of galaxies. Methods. We apply an object point process with interactions (the Bisous process) to trace and describe the filamentary network both in the observed samples (the 2dFGRS catalogue) and in the numerical models that have been prepared to mimic the data. We compare the networks. Results. We find that the properties of filaments in numerical models (mock samples) have a large variance. A few mock samples display filaments that resemble the observed filaments, but usually the model filaments are much shorter and do not form an extended network. Conclusions. We conclude that although we can build numerical models that are similar to observations in many respects, they may fail yet to explain the filamentary structure seen in the data. The Bisous-built filaments are a good test for such a structure.

show abstract

“…Methods that aim to reconstruct line networks (e.g., roads and hydrographic networks) from incomplete network delineations from remote sensing (Stoica et al, 2004;Lacoste et al, 2005) or vectorised cartographic elements (Baltsavias and Zhang, 2005;Mariani et al, 1995) have previously been proposed. The former method uses a random polyline stochastic simulation process derived from the Candy model (Descombes et al, 2001), and the latter uses a deterministic minimisation process (Kruskal, 1956;Prim, 1957).…”

Section: Introductionmentioning

confidence: 99%

A spatial stochastic algorithm to reconstruct artificial drainage networks from incomplete network delineations

Bailly

Levavasseur²,

Lagacherie³

2011

International Journal of Applied Earth Observation and Geoinfor

View full text Add to dashboard Cite

A spatial stochastic algorithm that aims to reconstruct an entire artificial drainage network of a cultivated landscape from disconnected reaches of the network is proposed here. This algorithm uses random network initialisation and a simulated annealing algorithm, both of which are based on random pruning or branching processes, to converge the multi-objective properties of the networks; the reconstructed networks are directed tree graphs, conform to a given cumulative length and maximise the proportion of reconnected reaches. This algorithm runs within a directed plot boundaries lattice, with the direction governed by elevation. The proposed algorithm was applied to a 2.6-km 2 catchment of a Languedocian vineyard in the south of France.The 24-km-long reconstructed networks maximised the reconnection of the reaches obtained either from a hydrographic database or remote sensing data processing. The distribution of the reconstructed networks compared to the actual networks was determined using specific topographical and topological metrics on the networks. The results show that adding data on disconnected * Corresponding author Email addresses: bailly@teledetection.fr (J.S. Bailly), levavasseur@supagro.inra.fr (F. Levavasseur), lagacherie@supagro.inra.fr (P. Lagacherie) Preprint submitted to International Journal of Applied Earth Observation and GeoinformationJune 7, 2011Author-produced version of the article published in International Journal of Applied Earth Observation and Geoinformation, 2011, 13(6) Doi : 10.1016Doi : 10. /j.jag.2011 reaches to constrain reconstruction, while increasing the accuracy of the reconstructed network topology, also adds biases to the geometry and topography of the reconstructed network. This network reconstruction method allows the mapping of uncertainties in the representation while integrating most of the available knowledge about the networks, including local data and global characteristics. It also permits the assessment of the benefits of the remote sensing partial detection process in drainage network mapping., 853-862. The original publication is available at http://www.sciencedirect.com/

show abstract

Point processes for unsupervised line network extraction in remote sensing

Cited by 142 publications

References 25 publications

Marked Point Process Model for Curvilinear Structures Extraction

Marked Point Process Model for Curvilinear Structures Extraction

Filaments in observed and mock galaxy catalogues

A spatial stochastic algorithm to reconstruct artificial drainage networks from incomplete network delineations

Contact Info

Product

Resources

About