遥感影像目标检测多尺度熵神经网络架构搜索
P237%TP751.1; 针对传统神经网络架构搜索需要耗费大量时间用于超网训练,搜索效率较低,搜索得到的模型无法高效解决遥感影像中多尺度目标检测困难、背景复杂度高的问题,本文提出采用多尺度熵神经网络架构搜索的方法进行遥感影像目标检测.首先,在搜索空间的基础模块中加入特征分离卷积以代替残差模块中的常规卷积,减少遥感影像中由于背景复杂度高而造成的信息间干扰,提高网络模型在复杂背景下的检测性能;然后,引入最大熵原理,计算搜索空间中每个候选网络的多尺度熵,将多尺度熵与特征金字塔网络相结合,以兼顾遥感影像大、中、小目标的检测;最后,在不进行参数训练的情况下利用渐进式进化算法搜索得到多尺度熵最大的网络模...
Saved in:
| Published in | 测绘学报 Vol. 53; no. 7; pp. 1384 - 1400 |
|---|---|
| Main Authors | , , , |
| Format | Journal Article |
| Language | Chinese |
| Published |
兰州交通大学电子与信息工程学院,甘肃兰州 730070%兰州交通大学测绘与地理信息学院,甘肃兰州 730070%挪威科技大学土木与环境工程学院,挪威特隆赫姆7491
12.08.2024
兰州交通大学测绘与地理信息学院,甘肃兰州 730070 |
| Subjects | |
| Online Access | Get full text |
| ISSN | 1001-1595 |
| DOI | 10.11947/j.AGCS.2024.20230455 |
Cover
| Abstract | P237%TP751.1; 针对传统神经网络架构搜索需要耗费大量时间用于超网训练,搜索效率较低,搜索得到的模型无法高效解决遥感影像中多尺度目标检测困难、背景复杂度高的问题,本文提出采用多尺度熵神经网络架构搜索的方法进行遥感影像目标检测.首先,在搜索空间的基础模块中加入特征分离卷积以代替残差模块中的常规卷积,减少遥感影像中由于背景复杂度高而造成的信息间干扰,提高网络模型在复杂背景下的检测性能;然后,引入最大熵原理,计算搜索空间中每个候选网络的多尺度熵,将多尺度熵与特征金字塔网络相结合,以兼顾遥感影像大、中、小目标的检测;最后,在不进行参数训练的情况下利用渐进式进化算法搜索得到多尺度熵最大的网络模型用于目标检测任务,在保证模型检测精度的同时,提升网络搜索效率.本文方法在RSOD、DIOR和DOTA数据集上的平均检测精度均值分别达到93.1％、75.5％和73.6％,网络搜索时间为8.1 h.试验结果表明,与当前基准方法相比,本文方法能够显著提升网络的搜索效率,在目标检测任务中更好地结合了不同尺度下的特征并解决了影像背景复杂度高的问题. |
|---|---|
| AbstractList | P237%TP751.1; 针对传统神经网络架构搜索需要耗费大量时间用于超网训练,搜索效率较低,搜索得到的模型无法高效解决遥感影像中多尺度目标检测困难、背景复杂度高的问题,本文提出采用多尺度熵神经网络架构搜索的方法进行遥感影像目标检测.首先,在搜索空间的基础模块中加入特征分离卷积以代替残差模块中的常规卷积,减少遥感影像中由于背景复杂度高而造成的信息间干扰,提高网络模型在复杂背景下的检测性能;然后,引入最大熵原理,计算搜索空间中每个候选网络的多尺度熵,将多尺度熵与特征金字塔网络相结合,以兼顾遥感影像大、中、小目标的检测;最后,在不进行参数训练的情况下利用渐进式进化算法搜索得到多尺度熵最大的网络模型用于目标检测任务,在保证模型检测精度的同时,提升网络搜索效率.本文方法在RSOD、DIOR和DOTA数据集上的平均检测精度均值分别达到93.1％、75.5％和73.6％,网络搜索时间为8.1 h.试验结果表明,与当前基准方法相比,本文方法能够显著提升网络的搜索效率,在目标检测任务中更好地结合了不同尺度下的特征并解决了影像背景复杂度高的问题. |
| Abstract_FL | Aiming at the traditional neural architecture search requires an enormous amount of time for supernet training,search efficiency is suboptimal,and the searched model can not efficiently solve the problem of multi-scale object detection dif-ficulty and high background complexity in remote sensing images.This paper proposes a multi-scale entropy neural architecture search method for object detection in remote sensing images.At first,the feature separation convolution is added to the base block of the search space instead of the regular convolution in the residual block,which reduces the information redundancy in remote sensing images due to the high background complexity,and improves the detection performance of the network mode-luner the complex background.Next,the maximum entropy principle is introduced to calculate the multi-scale entropy of each candidate network in the search space,and the multi-scale entropy is combined with the feature pyramid network to balance the detection of large,medium and small objects in remote sensing images.Finally,the network model with maximum multi-scale entropy is obtained by searching without parameter training using progressive evolutionary algorithm for the object detection task.The model ensures detection accuracy while improving the search efficiency.The proposed algorithm achieves a mean av-erage precision of 93.1％,75.5％and 73.6％on the RSOD,DIOR and DOTA datasets,respectively,with a network search time of 8.1 hours.The experimental results demonstrate that the proposed algorithm can significantly improve the search effi-ciency of the network,combine the features at different scales better and solve the problem of high image background complex-ity in the object detection task compared with the current benchmark methods. |
| Author | 闫浩文 解恒静 范红超 杨军 |
| AuthorAffiliation | 兰州交通大学测绘与地理信息学院,甘肃兰州 730070;兰州交通大学电子与信息工程学院,甘肃兰州 730070%兰州交通大学测绘与地理信息学院,甘肃兰州 730070%挪威科技大学土木与环境工程学院,挪威特隆赫姆7491 |
| AuthorAffiliation_xml | – name: 兰州交通大学测绘与地理信息学院,甘肃兰州 730070;兰州交通大学电子与信息工程学院,甘肃兰州 730070%兰州交通大学测绘与地理信息学院,甘肃兰州 730070%挪威科技大学土木与环境工程学院,挪威特隆赫姆7491 |
| Author_FL | FAN Hongchao YANG Jun XIE Hengjing YAN Haowen |
| Author_FL_xml | – sequence: 1 fullname: YANG Jun – sequence: 2 fullname: XIE Hengjing – sequence: 3 fullname: FAN Hongchao – sequence: 4 fullname: YAN Haowen |
| Author_xml | – sequence: 1 fullname: 杨军 – sequence: 2 fullname: 解恒静 – sequence: 3 fullname: 范红超 – sequence: 4 fullname: 闫浩文 |
| BookMark | eNotj71KA0EYRaeIYIx5BEvLXeebn8ws2IRFoxCwUOswM86qQTbgIloqBi0soq2GVdRYioKxSMjj7Ay-hQna3NtczuEuoFLaSS1CS4BDgIiJlXZYb8TbIcGEzYJixnkJlQFjCIBHfB5Vs-xQY8wZFZxGZbT6czFw3cdi8llc9vzDu3u6di_nbnhTvN4XH6Ni9Oavhn6Q-3HPT-78uO_yb5d33W3ffz0vorlEHWW2-t8VtLu-thNvBM2txmZcbwbZ1EwCCiYSStqaJIIYINJKq7Xdw5jWmIxASGZUQigzhFMjdaLAMuAajLBsOqIVtPzHPVVpotL9VrtzcpxOjS1zcKZnb7HAQOgv4PZgOw |
| ClassificationCodes | P237%TP751.1 |
| ContentType | Journal Article |
| Copyright | Copyright © Wanfang Data Co. Ltd. All Rights Reserved. |
| Copyright_xml | – notice: Copyright © Wanfang Data Co. Ltd. All Rights Reserved. |
| DBID | 2B. 4A8 92I 93N PSX TCJ |
| DOI | 10.11947/j.AGCS.2024.20230455 |
| DatabaseName | Wanfang Data Journals - Hong Kong WANFANG Data Centre Wanfang Data Journals 万方数据期刊 - 香港版 China Online Journals (COJ) China Online Journals (COJ) |
| DatabaseTitleList | |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Astronomy & Astrophysics |
| DocumentTitle_FL | Multi-scale entropy neural architecture search for object detection in re-mote sensing images |
| EndPage | 1400 |
| ExternalDocumentID | chxb202407012 |
| GroupedDBID | -01 2B. 4A8 5VS 5XA 5XB 7X2 92E 92I 93N ABJNI ACGFS AEUYN AFKRA ALMA_UNASSIGNED_HOLDINGS ATCPS BBNVY BENPR BHPHI BKSAR CCEZO CCPQU CCVFK CW9 HCIFZ IPNFZ M0K M7P OK1 P2P PATMY PCBAR PHGZM PHGZT PIMPY PMFND PSX PYCSY RIG TCJ TGP U1G U5K |
| ID | FETCH-LOGICAL-s1002-31c97a8e68272c128e8ebbed00364891784caf234c253c8bfa1e415b1c7e40033 |
| ISSN | 1001-1595 |
| IngestDate | Thu May 29 04:11:08 EDT 2025 |
| IsPeerReviewed | false |
| IsScholarly | true |
| Issue | 7 |
| Keywords | 遥感影像 最大熵 目标检测 maximum entropy multi-scale entropy 多尺度熵 渐进式进化 remote sensing object detection progressive evolution feature separation convolution 特征分离卷积 神经网络架构搜索 neural architecture search |
| Language | Chinese |
| LinkModel | OpenURL |
| MergedId | FETCHMERGED-LOGICAL-s1002-31c97a8e68272c128e8ebbed00364891784caf234c253c8bfa1e415b1c7e40033 |
| PageCount | 17 |
| ParticipantIDs | wanfang_journals_chxb202407012 |
| PublicationCentury | 2000 |
| PublicationDate | 2024-08-12 |
| PublicationDateYYYYMMDD | 2024-08-12 |
| PublicationDate_xml | – month: 08 year: 2024 text: 2024-08-12 day: 12 |
| PublicationDecade | 2020 |
| PublicationTitle | 测绘学报 |
| PublicationTitle_FL | Acta Geodaetica et Cartographica Sinica |
| PublicationYear | 2024 |
| Publisher | 兰州交通大学电子与信息工程学院,甘肃兰州 730070%兰州交通大学测绘与地理信息学院,甘肃兰州 730070%挪威科技大学土木与环境工程学院,挪威特隆赫姆7491 兰州交通大学测绘与地理信息学院,甘肃兰州 730070 |
| Publisher_xml | – name: 兰州交通大学测绘与地理信息学院,甘肃兰州 730070 – name: 兰州交通大学电子与信息工程学院,甘肃兰州 730070%兰州交通大学测绘与地理信息学院,甘肃兰州 730070%挪威科技大学土木与环境工程学院,挪威特隆赫姆7491 |
| SSID | ssib005437539 ssib038074662 ssib051373695 ssib002263888 ssib000862384 ssj0058465 |
| Score | 2.4392328 |
| Snippet | P237%TP751.1; 针对传统神经网络架构搜索需要耗费大量时间用于超网训练,搜索效率较低,搜索得到的模型无法高效解决遥感影像中多尺度目标检测困难、背景复杂度高的问题,本文... |
| SourceID | wanfang |
| SourceType | Aggregation Database |
| StartPage | 1384 |
| Title | 遥感影像目标检测多尺度熵神经网络架构搜索 |
| URI | https://d.wanfangdata.com.cn/periodical/chxb202407012 |
| Volume | 53 |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| journalDatabaseRights | – providerCode: PRVAFT databaseName: Open Access Digital Library issn: 1001-1595 databaseCode: KQ8 dateStart: 20120101 customDbUrl: isFulltext: true dateEnd: 99991231 titleUrlDefault: http://grweb.coalliance.org/oadl/oadl.html omitProxy: true ssIdentifier: ssib005437539 providerName: Colorado Alliance of Research Libraries |
| link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV3Na9RAFA_FXryIn9SvsojjRbZuJpPMDHiZbHctCoLYQm8lyWbtxS24LWhPikUPHqpXLVXUehQF66Glf84m-F_43st0E3XBWi_h7eTNm997b5J5b3Yy4ziXI0i3ZEekdQkDQl3IxKtHncSv86SRcrdDG5LgaovbwcycuDnvz4-NX6qsWlpZjqeS1ZHflRzGq1AGfsWvZP_Bs0OhUAA0-Beu4GG4HsjHrKUhEmTGZ62AKcF0m7V8Fk6z0EVCeUxBiWQ6ZKaFPKbBIPlHwsMlDkCEwBYis4Hqhqo3WFgQhpkAqytiAwIa0i0kwtBKhra0a0t0EwUCQxhYAiAh0aBbwCOY4dVouAKgkKAIyTS1CxoBAH-_R5AkuEMsgEiH5R3FjESVsI7LNEe7AK_WVRbVRINgQyCWYwngVBX5UAekhBaU0dQisMjqzAgXONXrlnk0wSGrockkGUhQG4Lc07BmhZ8AslQPOAWZlUrARlgLUBbV28y4Vh3TtpJN4QO177BCDqDWqA9vWploRODhpO3v2K7iwQGywbh_WNx_emyI20c_o0CJ-oCPRmry37gDdKUpwCmUi1aR1A-pz6jGKNwEDh-QAAlTxS3xasHtN38ge49CojQLNbEZMoGizlRIwEdJiuLcOBsJ4FpDiPX9aqhQ7GttX4myMu67XnHQoY0hXUHb746IT7SQFKCYG827U9hp8YKrFfwyIBsuk00WH8bIA_bFg9THOUQueDzNrTvVFAoSGFUN0WFErWxB6AtP-uXOv3i-hAjKLUJ915NeoIdTBpiP-LRCxKpvv3pE4NdGwaaPHXvdqHevEpfPHneO2YS6Zoq34wlnbHXxpDNh-vgX39L9R7UrNaKLGdz-Kef6jydb2drbwd7XwdP1_M3n7N3z7MPjbPvF4OPrwZedwc6n_Nl2vrWZ767ne6_y3Y1s83u2uZa93Mi_vT_tzLVbs82Zuj1Cpt7HvaUhwky0jFQaKC55ArF4qtI4Tju4DZdQ2pVKJFGXeyLhvpeouBu5KaQ0sZvIVOA5l2ecI72lXjrh1DTmYpCNSeVFEPnLmEcy0GlXqRRG464660xaQyzYIaK_8IsDz_2N4bxztHyNXXCOLD9YSS9CyrMcT5LPfwJxGOpy |
| linkProvider | Colorado Alliance of Research Libraries |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=%E9%81%A5%E6%84%9F%E5%BD%B1%E5%83%8F%E7%9B%AE%E6%A0%87%E6%A3%80%E6%B5%8B%E5%A4%9A%E5%B0%BA%E5%BA%A6%E7%86%B5%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%9E%B6%E6%9E%84%E6%90%9C%E7%B4%A2&rft.jtitle=%E6%B5%8B%E7%BB%98%E5%AD%A6%E6%8A%A5&rft.au=%E6%9D%A8%E5%86%9B&rft.au=%E8%A7%A3%E6%81%92%E9%9D%99&rft.au=%E8%8C%83%E7%BA%A2%E8%B6%85&rft.au=%E9%97%AB%E6%B5%A9%E6%96%87&rft.date=2024-08-12&rft.pub=%E5%85%B0%E5%B7%9E%E4%BA%A4%E9%80%9A%E5%A4%A7%E5%AD%A6%E7%94%B5%E5%AD%90%E4%B8%8E%E4%BF%A1%E6%81%AF%E5%B7%A5%E7%A8%8B%E5%AD%A6%E9%99%A2%2C%E7%94%98%E8%82%83%E5%85%B0%E5%B7%9E+730070%25%E5%85%B0%E5%B7%9E%E4%BA%A4%E9%80%9A%E5%A4%A7%E5%AD%A6%E6%B5%8B%E7%BB%98%E4%B8%8E%E5%9C%B0%E7%90%86%E4%BF%A1%E6%81%AF%E5%AD%A6%E9%99%A2%2C%E7%94%98%E8%82%83%E5%85%B0%E5%B7%9E+730070%25%E6%8C%AA%E5%A8%81%E7%A7%91%E6%8A%80%E5%A4%A7%E5%AD%A6%E5%9C%9F%E6%9C%A8%E4%B8%8E%E7%8E%AF%E5%A2%83%E5%B7%A5%E7%A8%8B%E5%AD%A6%E9%99%A2%2C%E6%8C%AA%E5%A8%81%E7%89%B9%E9%9A%86%E8%B5%AB%E5%A7%867491&rft.issn=1001-1595&rft.volume=53&rft.issue=7&rft.spage=1384&rft.epage=1400&rft_id=info:doi/10.11947%2Fj.AGCS.2024.20230455&rft.externalDocID=chxb202407012 |
| thumbnail_s | http://utb.summon.serialssolutions.com/2.0.0/image/custom?url=http%3A%2F%2Fwww.wanfangdata.com.cn%2Fimages%2FPeriodicalImages%2Fchxb%2Fchxb.jpg |