Graph-based semi-supervised learning with Local Binary Patterns for holistic object categorization

Dornaika, Fadi; Bosaghzadeh, Alireza; Salmane, Houssam; Ruichek, Yassine

doi:10.1016/j.eswa.2014.06.025

Cited by 19 publications

(6 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Shao et al [41] presented a graph by merging probabilistic class structure into the SR-based edge weighting model, which describes the edge weights between pixels and classes, enhancing the discriminability of graph structure. Dornaika et al [42] proposed graph-SSL with LBP for holistic object categorization, which designs a two-phase regularized least square to obtain graph in a semi-supervised context and utilizes LBP as descriptors to classify the detected objects in outdoor and indoor scenes. Cheng et al [43] developed a spectral-spatial random patches network, which integrates shallow, deep, spectral, spatial feature with LBP and stacks them into high dimensional vector.…”

Section: A Graph-based Ssl Hsi Classification Methodsmentioning

confidence: 99%

Hyperspectral Imagery Classification via Random Multigraphs Ensemble Learning

Miao

Chen

Yuan

et al. 2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Hyperspectral imagery (HSI) classification, which attempts to assign hyperspectral pixels with proper labels, has drawn significant attention in various applications. Recently, the graph-based semi-supervised learning (SSL) approaches have shown the outstanding ability to handle the situation of limited labeled data for classification task. However, it is hard to construct the pairwise adjacent graph due to the high dimensionality of hyperspectral data. Besides, the graph-based SSL models are usually decided by a single classifier, which fail to effectively learn the complex structures and intrinsic properties of HSI. To address these problems, we propose a novel graph-based SSL classification model for HSI, which is based on Random Multi-Graphs construction and Ensemble strategy (RMGE). Specifically, the anchor graph (AG) is constructed with spatial-spectral features, which integrates the spatial characteristics extracted by local binary pattern (LBP) on each selected spectrum, preserving fine structures of local region. In order to enhance the discriminative capability of the classifier and avoid the trivial solution, the maximum entropy regularization is added into adjacent AG model. In addition, to capture the diversity of HSI data effectively, we design the ensemble framework by employing multiple AGs to learn HSI features. Experiments conducted on real hyperspectral datasets indicate that the proposed RMGE shows better performance than that of state-of-the-art approaches.

show abstract

Section: A Graph-based Ssl Hsi Classification Methodsmentioning

confidence: 99%

Hyperspectral Imagery Classification via Random Multigraphs Ensemble Learning

Miao

Chen

Yuan

et al. 2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…As a future perspective, we are planning to test other combinations especially with a larger number of neighbors. Several studies [27,28,29,30,31,32] have presented the effectiveness of using LBP histograms (Figure 1) to learn graph-based classification models. Figure 2 shows the cases where the pattern can be uniform with only one change of the binary digits.…”

Section: Lbp and Uniform Lbp (Ulbp) Methodsmentioning

confidence: 99%

Deep Learning and Uniform LBP Histograms for Position Recognition of Elderly People with Privacy Preservation

Hamdi¹,

Bouhamed²,

Algarni³

et al. 2021

INT J COMPUT COMMUN, Int. J. Comput. Commun. Control

View full text Add to dashboard Cite

For the elderly population, falls are a vital health problem especially in the current context of home care for COVID-19 patients. Given the saturation of health structures, patients are quarantined, in order to prevent the spread of the disease. Therefore, it is highly desirable to have a dedicated monitoring system to adequately improve their independent living and significantly reduce assistance costs. A fall event is considered as a specific and brutal change of pose. Thus, human poses should be first identified in order to detect abnormal events. Prompted by the great results achieved by the deep neural networks, we proposed a new architecture for image classification based on local binary pattern (LBP) histograms for feature extraction. These features were then saved, instead of saving the whole image in the series of identified poses. We aimed to preserve privacy, which is highly recommended in health informatics. The novelty of this study lies in the recognition of individuals’ positions in video images avoiding the convolution neural networks (CNNs) exorbitant computational cost and Minimizing the number of necessary inputs when learning a recognition model. The obtained numerical results of our approach application are very promising compared to the results of using other complex architectures like the deep CNNs.

show abstract

“…Other popular options are SIFT features (Lowe, 2004), employed in works like Chang et al (2012), which show a relative invariance to translation, scale, rotation, illumination, or partial occlusion of objects, and SURF features, which are faster to retrieve and, as claimed in (Knopp et al, 2010), even more robust against those image transformations than SIFT features. A different, type of feature is such of Local Binary Pattern (LBP), which is fast to compute and describe the texture of a given portion of an image (Dornaika et al, 2014). Some recognition approaches have been built based on these primary features, like Mixture Models (Bourouis et al, 2014).…”

Section: Related Workmentioning

confidence: 99%