Use of Binary Partition Tree and energy minimization for object-based classification of urban land cover

Li, Mengmeng; Bijker, W.; Stein, Alfred

doi:10.1016/j.isprsjprs.2014.12.023

Cited by 26 publications

(31 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Future work could focus on analyzing the effect of building shadows on extracted roads, and could provide quantitative information to end-users whether a RoadROI should include building shadows or not. Moreover, because tree shadows affect building detection by means of shadow information (Li et al, 2015a), they are removed from the set of detected shadows that can be used for building detection. A Z-shape fuzzy membership function is used to quantify to which degree a pair of adjacent regions should be merged according to a geometrical feature.…”

Section: Discussionmentioning

confidence: 99%

“…She always takes a very good care of me, and always has my back. 4.1 Features used for classifying urban land cover (Li et al, 2015a …”

Section: Acknowledgmentsmentioning

confidence: 99%

“…The directional relationship is modeled by a fuzzy landscape of building shadows. The detailed description and computation of a fuzzy landscape can be found in Li et al (2015a). A reference (shadow) region R ref and a direction with angle β is considered.…”

Section: Obtain the Road Region Of Interestmentioning

confidence: 99%

“…For each image object, which is a segment from image segmentation, we calculated a set of spectral, geometrical, textual, and spatial features for classification (Table 4.1). Particularly, we calculated a directional feature based upon building shadow to improve the separability of building objects from other man-made objects (Li et al, 2015a(Li et al, , 2016a. A directional feature quantifies the directional relationship between building and shadow, where the directional relationship is modeled by a fuzzy landscape of building shadow.…”

Section: Urban Land Cover Classificationmentioning

confidence: 99%

“…We constructed two fuzzy rules for decision trees, based upon height and directional features. The detailed description on the directional feature can be found in Li et al (2015aLi et al ( , 2016a. This study uses hierarchical Bayesian models to determine the lower and upper bounds for these two fuzzy rules (van de Vlag and Stein, 2007) ( Figure 5.3c).…”

Section: Incorporating Osm For Urban Land Cover Classification By Baymentioning

confidence: 99%

See 4 more Smart Citations

Urban land use extraction from very high resolution remote sensing images

Self Cite

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

“…She always takes a very good care of me, and always has my back. 4.1 Features used for classifying urban land cover (Li et al, 2015a …”

Section: Acknowledgmentsmentioning

confidence: 99%

Section: Obtain the Road Region Of Interestmentioning

confidence: 99%

Section: Urban Land Cover Classificationmentioning

confidence: 99%

Section: Incorporating Osm For Urban Land Cover Classification By Baymentioning

confidence: 99%

See 3 more Smart Citations

Urban land use extraction from very high resolution remote sensing images

Self Cite

View full text Add to dashboard Cite

Optimized Hierarchical Cascaded Processing

Moons¹,

Bankman²,

Verhelst³

2018

Embedded Deep Learning

View full text Add to dashboard Cite

Recently, there has been an increasing demand for advanced classification capabilities embedded on wearable battery constrained devices, such as smartphones or -watches. Achieving such functionality with a tight power and energy budget has proven a real challenge, specifically for large-scale Neural Network based applications. Previously, cascaded systems have been proposed to minimize energy consumption for such applications, either through using a single wake-up stage, or by using a linear-or tree based cascade of consecutive classifiers that allow early termination. In this work, we expand upon these concepts by generalizing cascades to hierarchical cascaded processing, where a hierarchy of increasingly complex classifiers, each designed and trained for a specific subtask is used. This hierarchical approach significantly outperforms the wake-up based approach by up to 2 orders of magnitude in energy consumption at iso-accuracy, specifically in systems with sparse input data such as speech recognition and visual object detection. This paper presents a general design framework for such systems and illustrates how to optimize them towards minimum energy consumption. The text further proposes a roofline model for cascaded systems, derives system level trade-offs and proves the approaches validity through a visual classification case-study.

show abstract