As one of key tasks in the field of remote sensing image processing,object detection has always been a research hotspot.Although significant progresses have been made in this field,the deep learning methods still face significant challenges in dealing with scale changes and complex backgrounds in remote sensing images,which limits the further improvement of detection accuracy to some extent.To address this issue,an innovative object detection method for remote sensing images was proposed,which integrated a saliency guided image adaptive fusion module and improved Faster RCNN to enhance the accuracy of object detection.Firstly,in the image preprocessing stage,a saliency guided image adaptive fusion module was proposed,which effectively integrated the semantic information of the image and shallow fine-grained details,allowing the model to prioritize the object region while minimizing background interference.Secondly,after introducing MobileNetV3 as the feature extractor of Faster RCNN,an attention enhanced feature pyramid network was proposed,which combined attention with upsampling to further enhance target features and output high-quality feature maps,effectively improving the extraction effect of multi-dimensional features and providing more accurate and rich feature information for subsequent object detection tasks.Furthermore,a multi-scale region proposal network was designed,which can more accurately capture the features of objects of different sizes and shapes,thereby enhancing the expression ability of features and effectively improving the detection accuracy of targets.Finally,experiments on the DIOR and ROSD datasets demonstrated that the proposed network model exhibits higher detection accuracy compared to other advanced methods,fully demonstrating its superiority and effectiveness.
评价目标检测方法的性能指标有:AP(Average precision)表示每类平均精度,mAP(mean average precision)表示所有类的平均精度,mAP50表示预测边界框和真实边界框之间的交集与并集的比率(intersection over union,IoU)设置为0.5的平均精度,mAP50:95表示IoU设置为0.5到0.95的平均精度。mAP值越高检测性能越好。AP和mAP定义为
SharmaV K, MirR N.Saliency guided faster-RCNN (SGFr-RCNN) model for object detection and recognition[J].Journal of King Saud University-Computer and Information Sciences,2022,34(5):1687-1699.
[5]
ZhangG J, LuS J, ZhangW.CAD-net:a context-aware detection network for objects in remote sensing imagery[J].IEEE Transactions on Geoscience and Remote Sensing,2019,57(12):10015-10024.
[6]
HuangW, LiG Y, ChenQ Q,et al.CF2PN:a cross-scale feature fusion pyramid network based remote sensing target detection[J].Remote Sen-sing,2021,13(5):847.
[7]
LinT Y, DollárP, GirshickR,et al.Feature pyramid networks for object detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu:IEEE,2017:936-944.
[8]
LiZ Q, LiE Z, XuT Y,et al.Feature alignment FPN for oriented object detection in remote sensing images[J].IEEE Geoscience and Remote Sen-sing Letters,2023,20(2):1-5.
[9]
JiangX Z, WuY H.Remote sensing object detection based on convolution and swin transformer[J].IEEE Access,2023,11(3):38643-38656.
[10]
ZhangY, SongC L, ZhangD W.Small-scale aircraft detection in remote sensing images based on Faster-RCNN[J].Multimedia Tools and Applications,2022,81(13):18091-18103.
[11]
LuoM, TianY N, ZhangS W,et al.Individual tree detection in coal mine afforestation area based on improved faster RCNN in UAV RGB images[J].Remote Sensing,2022,14(21):5545.
[12]
PanY Y, ZhuN Z, DingL,et al.Identification and counting of sugarcane seedlings in the field using improved faster R-CNN[J].Remote Sensing,2022,14(22):5846.
[13]
KoonceB.MobileNetV3[M]//Convolutional Neural Networks with Swift for Tensorflow.Berkeley:Apress,2021:125-144.
[14]
JiY Z, ZhangH J, ZhangZ,et al.CNN-based encoder-decoder networks for salient object detection:a comprehensive review and recent advances[J].Information Sciences,2021,546(2):835-857.
[15]
DongY N, LiuQ W, DuB,et al.Weighted feature fusion of convolutional neural network and graph attention network for hyperspectral image classification[J].IEEE Transactions on Image Processing,2022,31:1559-1572.
[16]
ZhuW J, LiangS, WeiY C,et al.Saliency optimization from robust background detection[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition.Columbus:IEEE,2014:2814-2821.
[17]
RenS Q, HeK M, 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.
[18]
ChengG, HanJ W.A survey on object detection in optical remote sensing images[J].ISPRS Journal of Photogrammetry and Remote Sensing,2016,117(6):11-28.
[19]
LongY, GongY P, XiaoZ F,et al.Accurate object localization in remote sensing images based on convolutional neural networks[J].IEEE Transactions on Geoscience and Remote Sensing,2017,55(5):2486-2498.
[20]
GeZ, LiuS T, WangF,et al.YOLOX:exceeding YOLO series in 2021[EB/OL].(2021-08-06)[2023-08-15].