1.College of Information Science and Technology,Donghua University,Shanghai 201620,China
2.Engineering Research Center of Digitized Textile and Fashion Technology,Ministry of Education,Donghua University,Shanghai 201620,China
Show less
文章历史+
Received
Accepted
Published
2024-11-03
Issue Date
2026-04-29
PDF (2709K)
摘要
光学遥感图像目标检测是遥感图像数据智能解译的关键技术。为了解决遥感图像目标检测时,目标尺度差异大,目标受背景因素干扰,目标形状各异的问题,提出了LMK(large multiscale kernel)网络。该网络通过大核卷积分解和多尺度注意力机制模块,能够动态调整空间感受野,从而更好地捕获遥感场景中物体的上下文信息。此外,设计了一种面向目标检测的形状自适应选择(SAS,shape-adaptive selection)标签分配策略。该策略将目标形状信息集中于长宽比,通过结合物体的形状信息和特征分布计算IoU(intersection over union)最优阈值。针对遥感图像目标姿态旋转定位难的问题,引入了KFIoU损失函数。实验结果表明,所提出的目标检测模型在HRSC2016、UCAS-AOD和DOTA数据集上的精度分别达到了96.73%、97.85%和77.26%。改进后的模型优于目前绝大多数目标检测算法。
Abstract
Target detection in optical remote sensing images is a key technology for the intelligent interpretation of remote sensing data. To address the challenges posed by of significant scale variations, background interference and the diversity of target shapes in remote sensing image detection, this paper proposes the Large Multi-scale Kernel (LMK) network. This network utilized large kernel convolution and a multi-scale attention mechanism to dynamically adjust the spatial receptive field, thereby enhancing the capture of contextual information about objects in remote sensing scenes. Furthermore, a Shape-Adaptive Selection (SAS) label allocation strategy was designed for target detection, which focused on the aspect ratio of the target shapes and calculated an optimal IoU threshold based on the shape information and feature distribution. To address the difficulty of target orientation and positioning in remote sensing images, this paper introduced the KFIoU loss function. Experimental results show that the proposed target detection model achieves accuracies of 96.73%, 97.85%, and 77.26% on the HRSC 2016, UCAS-AOD, and DOTA datasets, respectively, outperforming most existing target detection algorithms.
CHENGGong, HANJunwei. A survey on object detection in optical remote sensing images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 117: 11-28.
RENShaqing, HEKaiming, GIRSHICKR, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[4]
LIUZhuang, MAOHanzi, WUChaoyuan, et al. A ConvNet for the 2020s[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA:IEEE, 2022: 11966-11976.
[5]
DINGXiaohan, ZHANGXiangyu, HANJungong, et al. Scaling up your kernels to 31 × 31: Revisiting large kernel design in CNNs[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA:IEEE, 2022: 11953-11965.
[6]
LIUShiwei, CHENTianlong, CHENXiaohan, et al. More ConvNets in the 2020s: Scaling up kernels beyond 51×51 using sparsity[EB/OL].(2023-03-03)[2024-11-03].
GUOMenghao, LUChengze, HOUQibin, et al. Segnext: Rethinking convolutional attention design for semantic segmentation[J]. Advances in Neural Information Processing Systems, 2022, 35: 1140-1156.
[9]
HOUQibin, LUChengze, CHENGMingming, et al. Conv2Former: A simple transformer-style ConvNet for visual recognition[EB/OL]. (2022-11-22)[2024-11-03].
[10]
YANGTong, ZHANGXianyu, LIZeming, et al. MetaAnchor: Learning to detect objects with customized anchors[EB/OL].(2018-07-03)[2024-11-03].
[11]
MINGQi, ZHOUZhiqiang, MIAOLingjuan, et al. Dynamic anchor learning for arbitrary-oriented object detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(3): 2355-2363.
[12]
TIANZhi, SHENChunhua, CHENHao, et al. FCOS: Fully convolutional one-stage object detection[C]// IEEE/CVF International Conference on Computer Vision (ICCV). Republic of Korea: IEEE, 2020:9626-9635.DOI:10.1109/ICCV.2019.00962 .
[13]
KIMK, LEEH S. Probabilistic anchor assignment with IoU prediction for object detection[M]// VEDALDI A, BISCHOF H, BROX T, et al. Lecture Notes in Computer Science. Cham: Springer International Publishing, 2020: 355-371.
[14]
LIYuxuan, HOUQibin, ZHENGZhaohui, et al. Large selective kernel network for remote sensing object detection[C]//IEEE/CVF International Conference on Computer Vision (ICCV). Paris, France:IEEE, 2023: 16748-16759.
[15]
YANGXue, ZHOUYue, ZHANGGefan, et al. The KFIoU loss for rotated object detection[EB/OL].(2023-02-06)[2024-11-03].
[16]
ZHANGShifeng, CHICheng, YAOYongqiang, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA:IEEE, 2020: 9759-9768.
[17]
YANGXue, YANJunchi, FENGZiming, et al. R3Det: Refined single-stage detector with feature refinement for rotating object[J].Proceedings of the AAAI Conference on Artificial Intelligence,2021, 35(4): 3163-3171.
[18]
QIANWen, YANGXue, PENGSilong, et al. RSDet: Point-based modulated loss for more accurate rotated object detection[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(11): 7869-7879.
[19]
ZHAOPengbo, QUZhenshen, BUYingjia, et al. PolarDet: A fast, more precise detector for rotated target in aerial images[J]. International Journal of Remote Sensing, 2021, 42(15): 5831-5861.
[20]
YANGXue, HOULiping, ZHOUYue, et al. Dense label encoding for boundary discontinuity free rotation detection[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN,USA:IEEE,2021:15819-15829.
[21]
HANJiaming, DINGJian, LIJie, et al. Align deep features for oriented object detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 5602511.DOI: 10.1109/TGRS.2021.3111234 .
[22]
CHENGGong, WANGJiabo, LIKe, et al. Anchor-free oriented proposal generator for object detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5625411.DOI: 10.1109/TGRS.2022.3183022 .
[23]
PANXingjia, RENYuqiang, SHENGKekai, et al. Dynamic refinement network for oriented and densely packed object detection[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA:IEEE, 2020: 11207-11216.
[24]
HOULiping, LUKe, XUEJian, et al. Shape-adaptive selection and measurement for oriented object detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36(1): 923-932.