Improved remote sensing image target detection based on YOLOv7

Abstract

Remote sensing images are taken at high altitude from above, with complex spatial scenes of images and a large number of target types. The detection of image targets on large scale remote sensing images suffers from the problem of small target size and target density. This paper proposes an improved model for remote sensing image detection based on you only look once version 7 (YOLOv7). First, the small-scale detection layer is added to reacquire tracking frames to improve the network’s recognition ability of small-scale targets, and then Bottleneck Transformers are fused in the backbone to make full use of the convolutional neural network (CNN)+Transformer architecture to enhance the feature extraction ability of the network. After that, the convolutional block attention module (CBAM) mechanism is added in the head to improve the model’s ability of small-scale target. Finally, the non-maximum suppressed (NMS) of YOLOv7 algorithm is changed to distance intersection over union-non maximum suppression (DIOU-NMS) to improve the detection ability of overlapping targets in the network. The results show that the method in this paper can improve the detection rate of small-scale targets in remote sensing images and effectively solve the problem of high overlap and is tested on the NWPU-VHR10 and DOTA1.0 datasets, and the accuracy of the improved model is improved by 6.3% and 4.2%, respectively, compared with the standard YOLOv7 algorithm.

References

[1] LuX, ZhengX, YuanY. Remote sensing scene classification by unsupervised representation learning[J]. IEEE transactions on geoscience and remote sensing, 2017, 55(9):5148-5157

[2] AfaqY, ManochaA. Analysis on change detection techniques for remote sensing applications: a review[J]. Ecological informatics, 2021, 63: 101310

[3] ZhaoZ Q, ZhengP, XuS, et al.. Object detection with deep learning: a review[J]. IEEE transactions on neural networks and learning systems, 2019, 30(11):3212-3232

[4] ShafiqueA, CaoG, KhanZ, et al.. Deep learning-based change detection in remote sensing images: a review[J]. Remote sensing, 2022, 14(4):871

[5] GirshickR, DonahueJ, DarrellT, et al.. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, June 23–28, 2014, Columbus, OH, USA. New York: IEEE, 2014, 978: 580-587

[6] RenS Q, HeK, GirshickR, et al.. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis & machine intelligence, 2017, 39(06):1137-1149

[7] LiuW, AnguelovD, ErhanD, et al.. SSD: single shot multibox detector[C], 2016, Berlin, Heidelberg, Springer International Publishing: 21-37

[8] RedmonJ, DivvalaS, GirshickR, et al.. You only look once: unified, real-time object detection[C], 2016, New York, IEEE: 779-788

[9] RedmonJ, FarhadiA. Yolo9000: better, faster, stronger[C], 2017, New York, IEEE: 7263-7271

[10] REDMON J, FARHADI A. Yolov3: an incremental improvement[EB/OL]. (2018-04-08) [2023-01-23]. https://arxiv.org/abs/1804.02767.

[11] ZHANG H, CISSE M, DAUPHIN Y N, et al. Mixup: beyond empirical risk minimization[EB/OL]. (2017-10-25) [2023-01-23]. https://arxiv.org/abs/1710.09412.

[12] WangC, ShiJ, YangX, et al.. Geospatial object detection via deconvolutional region proposal network[J]. IEEE journal of selected topics in applied earth observations and remote sensing, 2019, 12(8):3014-3027

[13] ChengG, ZhouP, HanJ. Learning rotation-invariant convolutional neural networks for object detection in VHR optical remote sensing images[J]. IEEE transactions on geoscience and remote sensing, 2016, 54(12):7405-7415

[14] YuX, GongY, JiangN, et al.. Scale match for tiny person detection[C], 2020, New York, IEEE: 1257-1265

[15] LuoH, WangP, ChenH, et al.. Object detection method based on shallow feature fusion and semantic information enhancement[J]. IEEE sensors journal, 2021, 21(19):21839-21851

[16] DengC, WangM, LiuL, et al.. Extended feature pyramid network for small object detection[J]. IEEE transactions on multimedia, 2021, 24: 1968-1979

[17] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. Yolov4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23) [2023-01-23]. https://arxiv.org/abs/2004.10934.

[18] WangC Y, BochkovskiyA, LiaoH Y M. Yolov7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C], 2023, New York, IEEE: 7464-7475

[19] SrinivasA, LinT Y, ParmarN, et al.. Bottleneck transformers for visual recognition[C], 2021, New York, IEEE: 16514-16524

[20] WooS, ParkJ, LeeJ Y, et al.. CBAM: convolutional block attention module[C], 2018, Berlin, Heidelberg, Springer International Publishing: 3-19

[21] ZHENG Z, WANG P, LIU W, et al. Distance-iouloss: faster and better learning for bounding box regression[EB/OL]. (2019-11-19) [2023-01-23]. https://arxiv.org/abs/1911.08287.

[22] ChengG, HanJ, ZhouP, et al.. Multi-class geospatial object detection and geographic image classification based on collection of part detectors[J]. ISPRS Journal of photogrammetry and remote sensing, 2014, 986: 119-132

[23] LiK, ChengG, BuS, et al.. Rotation-insensitive and context-augmented object detection in remote sensing images[J]. IEEE transactions on geoscience and remote sensing, 2017, 56(4):2337-2348

[24] YangX, YangJ, YanJ, et al.. SCRDet: towards more robust detection for small, cluttered and rotated objects[C], 2019, New York, IEEE: 8231-8240

[25] LiC, XuC, CuiZ, et al.. Feature-attentioned object detection in remote sensing imagery[C], 2019, New York, IEEE: 3886-3890978

[26] QIAN W, YANG X, PENG S, et al. Learning modulated loss for rotated object detection[EB/OL]. (2019-11-19) [2023-01-23]. https://arxiv.org/abs/1911.08299.