Computationally Efficient Mean-Shift Parallel Segmentation Algorithm for High-Resolution Remote Sensing Images

Wu, Tianjun; Xia, Liegang; Luo, Jiancheng; Zhou, Xiaocheng; Hu, Xiaodong; Ma, Jianghong; Song, Xueli

doi:10.1007/s12524-018-0841-8

Cited by 6 publications

(5 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In column (1), the classifications of "Low Vegetation" and "Clutter/background" were incorrectly mixed due to the similarity of their colors. In column (2), the DeepLabV3+ model incorrectly split "Tree" into "Low Vegetation" and "Clutter/Background." The Swin-ViT model correctly classified these, but the area was incomplete.…”

Section: Comparative Experimentsmentioning

confidence: 99%

“…They contain a rich amount of information on the texture, shape, structure, and neighborhood relationship of various features. The traditional mathematical theory-based semantic segmentation methods [1][2][3] for the remote sensing of images can be used for relatively simple contents, but are often not suitable for images with complex features. With the excellent image feature extraction capability shown by CNN in recent years, an end-to-end network structure has been established for use in image classification, semantic segmentation, object detection, and other fields, and is effectively used for remote sensing applications [4][5][6].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

ACTNet: A Dual-Attention Adapter with a CNN-Transformer Network for the Semantic Segmentation of Remote Sensing Imagery

Zhang

Liu

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

In recent years, the application of semantic segmentation methods based on the remote sensing of images has become increasingly prevalent across a diverse range of domains, including but not limited to forest detection, water body detection, urban rail transportation planning, and building extraction. With the incorporation of the Transformer model into computer vision, the efficacy and accuracy of these algorithms have been significantly enhanced. Nevertheless, the Transformer model’s high computational complexity and dependence on a pre-training weight of large datasets leads to a slow convergence during the training for remote sensing segmentation tasks. Motivated by the success of the adapter module in the field of natural language processing, this paper presents a novel adapter module (ResAttn) for improving the model training speed for remote sensing segmentation. The ResAttn adopts a dual-attention structure in order to capture the interdependencies between sets of features, thereby improving its global modeling capabilities, and introduces a Swin Transformer-like down-sampling method to reduce information loss and retain the original architecture while reducing the resolution. In addition, the existing Transformer model is limited in its ability to capture local high-frequency information, which can lead to an inadequate extraction of edge and texture features. To address these issues, this paper proposes a Local Feature Extractor (LFE) module, which is based on a convolutional neural network (CNN), and incorporates multi-scale feature extraction and residual structure to effectively overcome this limitation. Further, a mask-based segmentation method is employed and a residual-enhanced deformable attention block (Deformer Block) is incorporated to improve the small target segmentation accuracy. Finally, a sufficient number of experiments were performed on the ISPRS Potsdam datasets. The experimental results demonstrate the superior performance of the model described in this paper.

show abstract

Section: Comparative Experimentsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

ACTNet: A Dual-Attention Adapter with a CNN-Transformer Network for the Semantic Segmentation of Remote Sensing Imagery

Zhang

Liu

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…A utilização do algoritmo Mean Shift no contexto da análise de previsão de sistemas eólicos implica em uma abordagem diferenciada para provisão em séries temporais (Wu et al, 2018). Logo, admitindo uma amostra de dados x 0 = (x 1 ) n i=1 ER D , com base no modelo não-paramétrico de janela de Parzen, a função de densidade de probabilidade é definida por:…”

Section: Algoritmo Mean Shiftunclassified

“…O deslocamento entre os dois objetivos pode ser otimizado e representado pela fórmula de otimização irrestrita sob o parâmetro λ (Lima et al, 2013) (Wu et al, 2018).…”

Section: Algoritmo Mean Shiftunclassified

MeanShift para Densificação de Dados Aplicado à Previsão de Geração de Energia Eólica

Hellen Dianne P. de Souza,

Shigeaki Leite de Lima,

Denisson Queiroz Oliveira

2022

Proceedings IX Simpósio Brasileiro De Sistemas Elétricos

View full text Add to dashboard Cite

The forecast of wind energy generation is used by many energy companies in order to safely estimate the wind resource and in this way guarantee both the company and the consumer the increase in revenue and the supply of energy from a renewable source over a period of time. Some techniques were developed for this purpose and allow estimating the wind resource and the energy generated for a few hours, days or even weeks. In this work, an information theoretic learning technique for the densification of a wind measurement database is presented, generating new data with the same probability density function of the original set until obtaining the mode of the set, using these virtual data as scenarios of the forecast. Resumo: A previsão de geração de energia eólica é usada por muitas empresas de energia com a finalidade de estimar com segurança o recurso eólico, e desta forma garantir tanto para a empresa quanto ao consumidor o aumento da receita e o fornecimento de energia de fonte renovável durante um período de tempo. Muitas técnicas foram desenvolvidas com este objetivo e permitem estimar o recurso eólico e a energia gerada desde algumas horas, dias ou mesmo semanas. Neste trabalho é apresentado uma técnica baseada em teoria da informação para a densificação de uma base de dados de medições de vento, gerando novos dados com a mesma função densidade de probabilidade do conjunto original até obter a moda do conjunto, utilizando estes dados virtuais como cenários de previsão.

show abstract

“…However, two segments that have a small contact surface may belong to the same ground object, and two segments that have a large contact surface may belong to different ground objects. In [15], the authors present a parallel implementation of the mean-shift segmentation algorithm. They use a novel buffer-zone-based data-partitioning strategy to avoid the inconsistency on the boundaries of adjacent data chunks.…”

Section: Introductionmentioning

confidence: 99%

Large Remote Sensing Image Segmentation with Stitching Strategy Based on Dominant Color

Zhang¹,

Li-gang²,

Tong³

2021

JSW

View full text Add to dashboard Cite

Large remote sensing image segmentation is a crucial issue in object-based image analysis. It is common sense that a segmentation framework consists of three components: (1) dividing largeremote sensing image into blocks for overcoming the constraint of computer memory; (2) executing segmentation algorithm for each block individually; (3) stitching segmentation results of all blocks into a complete result for eliminating artificial borderscreated by dividing blocks. However, there is a lack of mature technologies to eliminate artificial borders produced by dividing blocks. In this paper, we proposed a new stitching strategy based on the dominant color similarity measure and modified thetraditional methodof dominant color similarity measure to make itmoresuitable for measuring the similarity of two segmented regions. A multi-scale segmentation algorithm is adopted for segmenting each block. External memory is used to store intermediate segmentation results and exchange data with internal memory. We tested the algorithm with three different images and validated that the algorithm can implement the segmentation for large remote sensing images in a common computer. Experiments demonstrate that the stitchingstrategy based on the similarity measure of dominant color can effectively eliminate artificial borders.

show abstract

Computationally Efficient Mean-Shift Parallel Segmentation Algorithm for High-Resolution Remote Sensing Images

Cited by 6 publications

References 19 publications

ACTNet: A Dual-Attention Adapter with a CNN-Transformer Network for the Semantic Segmentation of Remote Sensing Imagery

ACTNet: A Dual-Attention Adapter with a CNN-Transformer Network for the Semantic Segmentation of Remote Sensing Imagery

MeanShift para Densificação de Dados Aplicado à Previsão de Geração de Energia Eólica

Large Remote Sensing Image Segmentation with Stitching Strategy Based on Dominant Color

Contact Info

Product

Resources

About