A Multi-Objective Semantic Segmentation Algorithm Based on Improved U-Net Networks

Hao, Xuejie; Yin, Lizeyan; Li, Xiuhong; Zhang, Le; Yang, Rongjin

doi:10.3390/rs15071838

Cited by 14 publications

(12 citation statements)

References 24 publications

(32 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, these methods encounter certain limitations highlighted by [16]: (1) The alignment of segmentations or aircraft often necessitates direction estimation, which is constrained by the methods' capacity and may result in inaccuracies in direction prediction. (2) The coarse shape representations employed by these techniques tend to overlook crucial details that are essential for discriminating between different aircraft types.…”

Section: Classical Methods For Aircraft Recognitionmentioning

confidence: 99%

“…This task involves assigning a class label to each pixel within an image. Over the past ten years, advancements in technology, such as compact imaging sensors and unmanned aerial vehicles, have yielded significant enhancements in image quality while reducing operational and equipment costs [2]. These cost-effective platforms provide flexible access to both multi-spectral and high-resolution images, along with an accelerated data acquisition rate [3].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Remote Sensing Image Segmentation for Aircraft Recognition Using U-Net as Deep Learning Architecture

Shaar,

Yılmaz,

Topcu

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

Recognizing aircraft automatically by using satellite images has different applications in both the civil and military sectors. However, due to the complexity and variety of the foreground and background of the analyzed images, it remains challenging to obtain a suitable representation of aircraft for identification. Many studies and solutions have been presented in the literature, but only a few studies have suggested handling the issue using semantic image segmentation techniques due to the lack of publicly labeled datasets. With the advancement of CNNs, researchers have presented some CNN architectures, such as U-Net, which has the ability to obtain very good performance using a small training dataset. The U-Net architecture has received much attention for segmenting 2D and 3D biomedical images and has been recognized to be highly successful for pixel-wise satellite image classification. In this paper, we propose a binary image segmentation model to recognize aircraft by exploiting and adopting the U-Net architecture for remote sensing satellite images. The proposed model does not require a significant amount of labeled data and alleviates the need for manual aircraft feature extraction. The public dense labeling remote sensing dataset is used to perform the experiments and measure the robustness and performance of the proposed model. The mean IoU and pixel accuracy are adopted as metrics to assess the obtained results. The results in the testing dataset indicate that the proposed model can achieve a 95.08% mean IoU and a pixel accuracy of 98.24%.

show abstract

Section: Classical Methods For Aircraft Recognitionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Remote Sensing Image Segmentation for Aircraft Recognition Using U-Net as Deep Learning Architecture

Shaar,

Yılmaz,

Topcu

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…Ref. [28] added a csAG module consisting of spatial attention and channel attention to the HRNetv2 [11] network to minimize the error rate. The existing algorithms have taken advantage of the high-resolution representation of HRNet.…”

Section: Multi-resolution Feature Learningmentioning

confidence: 99%

High-Resolution Network with Transformer Embedding Parallel Detection for Small Object Detection in Optical Remote Sensing Images

Zhang,

Liu,

Chang

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Small object detection in remote sensing enables the identification and analysis of unapparent but important information, playing a crucial role in various ground monitoring tasks. Due to the small size, the available feature information contained in small objects is very limited, making them more easily buried by the complex background. As one of the research hotspots in remote sensing, although many breakthroughs have been made, there still exist two significant shortcomings for the existing approaches: first, the down-sampling operation commonly used for feature extraction can barely preserve weak features of objects in a tiny size; second, the convolutional neural network methods have limitations in modeling global context to address cluttered backgrounds. To tackle these issues, a high-resolution network with transformer embedding parallel detection (HRTP-Net) is proposed in this paper. A high-resolution feature fusion network (HR-FFN) is designed to solve the first problem by maintaining high spatial resolution features with enhanced semantic information. Furthermore, a Swin-transformer-based mixed attention module (STMA) is proposed to augment the object information in the transformer block by establishing a pixel-level correlation, thereby enabling global background–object modeling, which can address the second shortcoming. Finally, a parallel detection structure for remote sensing is constructed by integrating the attentional outputs of STMA with standard convolutional features. The proposed method effectively mitigates the impact of the intricate background on small objects. The comprehensive experiment results on three representative remote sensing datasets with small objects (MASATI, VEDAI and DOTA datasets) demonstrate that the proposed HRTP-Net achieves a promising and competitive performance.

show abstract

“…Noticeably, a comprehensive transportation hub represents a crucial aspect of the national or regional transportation system, is the intersection of several modes of transportation network, and is by several kinds of transportation connected to the fixed and mobile equipment that comprise the whole [4], in which the productive activity of transportation is to achieve the displacement of people, goods, and information transmission, mainly including that of highway, rail transit, maritime, and air transport. Besides this, the advancement of transportation infrastructure plays a crucial role in enhancing people's quality of life, stimulating economic progress, upholding social welfare, and bolstering national security [5]. Further, transportation accessibility represents and influences regional socio-economic development to a large extent.…”

Section: Introductionmentioning

confidence: 99%

Integration of Building Information Modeling (BIM) with Transportation and Facilities: Recent Applications and Future Perspectives

Liu,

Deng,

Liu

et al. 2024

Buildings

View full text Add to dashboard Cite

At present, increased modes of transport have facilitated daily life. Building information modeling (BIM) integration has become a key strategy to foster efficiency, collaboration, and sustainability in the fields of buildings, transport, and facilities. Currently, there is a scarcity of comprehensive examinations of the performance of integrated BIM with transport facilities and applications to help identify potential prospects for sustainable development. Hence, this research paper attempts to scrutinize the integration of BIM, transportation, and facilities (T and Fs) by highlighting future directions and trends, revealing the existing status and hotspots of research in the field, and clarifying the developmental pulse of research as well as emerging areas in the future. A quantitative research technique is utilized in this study, using VOSviewer and CiteSpace software, and the Web of Science Core Collection (WoSCC) database. The study findings suggest that the hot keywords for the integration of BIM and T and Fs are predominantly concentrated on construction, framework, system, design, and management. In addition, over the previous 34 years (years 1989 to 2023), the applications of point cloud, digital twin, and life cycle assessment have been the current hot topics, and these emerging technologies can offer more innovative breakthroughs for the future. Furthermore, the integration of BIM and T and Fs represents an important trend that is essential for improving the efficiency, sustainability, and intelligence of buildings and infrastructure, from which the three domains can create synergies that contribute to the better planning, construction, and management of building and infrastructure projects.

show abstract

A Multi-Objective Semantic Segmentation Algorithm Based on Improved U-Net Networks

Cited by 14 publications

References 24 publications

Remote Sensing Image Segmentation for Aircraft Recognition Using U-Net as Deep Learning Architecture

Remote Sensing Image Segmentation for Aircraft Recognition Using U-Net as Deep Learning Architecture

High-Resolution Network with Transformer Embedding Parallel Detection for Small Object Detection in Optical Remote Sensing Images

Integration of Building Information Modeling (BIM) with Transportation and Facilities: Recent Applications and Future Perspectives

Contact Info

Product

Resources

About