{"title":"Frontiers | Remote sensing object detection with feature-associated convolutional neural networks","authors":"Jianghao Rao, Tao Wu, Hongyun Li, Jianlin Zhang, Qiliang Bao, Zhenming Peng","doi":"10.3389/feart.2024.1381192","DOIUrl":null,"url":null,"abstract":"Neural networks have become integral to remote sensing data processing. Among neural networks, convolutional neural networks (CNNs) in deep learning offer numerous advanced algorithms for object detection in remote sensing imagery, which is pivotal in military and civilian contexts. CNNs excel in extracting features from training samples. However, traditional CNN models often lack specific signal assumptions tailored to remote sensing data at the feature level. In this paper, we propose a novel approach aimed at effectively representing and correlating information within CNNs for remote sensing object detection. We introduce object tokens and incorporate global information features in embedding layers, facilitating the comprehensive utilization of features across multiple hierarchical levels. Consideration of feature maps from images as two-dimensional signals, matrix image signal processing is employed to correlate features for diverse representations within the CNN framework. Moreover, hierarchical feature signals are effectively represented and associated during end-to-end network training. Experiments on various datasets demonstrate that the CNN model incorporating feature representation and association outperforms CNN models lacking these elements in object detection from remote sensing images. Additionally, integrating image signal processing enhances efficiency in end-to-end network training. Various signal processing approaches increase the process ability of the network, and the methodology could be transferred to other specific and well-defined task.","PeriodicalId":12359,"journal":{"name":"Frontiers in Earth Science","volume":null,"pages":null},"PeriodicalIF":2.0000,"publicationDate":"2024-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Frontiers in Earth Science","FirstCategoryId":"89","ListUrlMain":"https://doi.org/10.3389/feart.2024.1381192","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"GEOSCIENCES, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
Neural networks have become integral to remote sensing data processing. Among neural networks, convolutional neural networks (CNNs) in deep learning offer numerous advanced algorithms for object detection in remote sensing imagery, which is pivotal in military and civilian contexts. CNNs excel in extracting features from training samples. However, traditional CNN models often lack specific signal assumptions tailored to remote sensing data at the feature level. In this paper, we propose a novel approach aimed at effectively representing and correlating information within CNNs for remote sensing object detection. We introduce object tokens and incorporate global information features in embedding layers, facilitating the comprehensive utilization of features across multiple hierarchical levels. Consideration of feature maps from images as two-dimensional signals, matrix image signal processing is employed to correlate features for diverse representations within the CNN framework. Moreover, hierarchical feature signals are effectively represented and associated during end-to-end network training. Experiments on various datasets demonstrate that the CNN model incorporating feature representation and association outperforms CNN models lacking these elements in object detection from remote sensing images. Additionally, integrating image signal processing enhances efficiency in end-to-end network training. Various signal processing approaches increase the process ability of the network, and the methodology could be transferred to other specific and well-defined task.
期刊介绍:
Frontiers in Earth Science is an open-access journal that aims to bring together and publish on a single platform the best research dedicated to our planet.
This platform hosts the rapidly growing and continuously expanding domains in Earth Science, involving the lithosphere (including the geosciences spectrum), the hydrosphere (including marine geosciences and hydrology, complementing the existing Frontiers journal on Marine Science) and the atmosphere (including meteorology and climatology). As such, Frontiers in Earth Science focuses on the countless processes operating within and among the major spheres constituting our planet. In turn, the understanding of these processes provides the theoretical background to better use the available resources and to face the major environmental challenges (including earthquakes, tsunamis, eruptions, floods, landslides, climate changes, extreme meteorological events): this is where interdependent processes meet, requiring a holistic view to better live on and with our planet.
The journal welcomes outstanding contributions in any domain of Earth Science.
The open-access model developed by Frontiers offers a fast, efficient, timely and dynamic alternative to traditional publication formats. The journal has 20 specialty sections at the first tier, each acting as an independent journal with a full editorial board. The traditional peer-review process is adapted to guarantee fairness and efficiency using a thorough paperless process, with real-time author-reviewer-editor interactions, collaborative reviewer mandates to maximize quality, and reviewer disclosure after article acceptance. While maintaining a rigorous peer-review, this system allows for a process whereby accepted articles are published online on average 90 days after submission.
General Commentary articles as well as Book Reviews in Frontiers in Earth Science are only accepted upon invitation.