High Quality Object Detection for Multiresolution Remote Sensing Imagery Using Cascaded Multi-Stage Detectors

被引:5
作者
Wu, Binglong [1 ,2 ]
Shen, Yuan [1 ]
Guo, Shanxin [1 ,3 ]
Chen, Jinsong [1 ,3 ]
Sun, Luyi [1 ,3 ]
Li, Hongzhong [1 ,3 ]
Ao, Yong [2 ]
机构
[1] Chinese Acad Sci, Shenzhen Inst Adv Technol, Ctr Geospatial Informat, Shenzhen 518055, Peoples R China
[2] Changan Univ, Sch Earth Sci & Resources, 126 Yanta Rd, Xian 710054, Peoples R China
[3] Shenzhen Engn Lab Ocean Environm Big Data Anal &, Shenzhen 518055, Peoples R China
基金
中国国家自然科学基金;
关键词
object detection; cascaded detectors; Intersection over Union (IoU) threshold; classification ensemble; bounding box regression; multiresolution remote sensing images; NETWORK; TEMPLATE;
D O I
10.3390/rs14092091
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
Deep-learning-based object detectors have substantially improved state-of-the-art object detection in remote sensing images in terms of precision and degree of automation. Nevertheless, the large variation of the object scales makes it difficult to achieve high-quality detection across multiresolution remote sensing images, where the quality is defined by the Intersection over Union (IoU) threshold used in training. In addition, the imbalance between the positive and negative samples across multiresolution images worsens the detection precision. Recently, it was found that a Cascade region-based convolutional neural network (R-CNN) can potentially achieve a higher quality of detection by introducing a cascaded three-stage structure using progressively improved IoU thresholds. However, the performance of Cascade R-CNN degraded when the fourth stage was added. We investigated the cause and found that the mismatch between the ROI features and the classifier could be responsible for the degradation of performance. Herein, we propose a Cascade R-CNN++ structure to address this issue and extend the three-stage architecture to multiple stages for general use. Specifically, for cascaded classification, we propose a new ensemble strategy for the classifier and region of interest (RoI) features to improve classification accuracy at inference. In localization, we modified the loss function of the bounding box regressor to obtain higher sensitivity around zero. Experiments on the DOTA dataset demonstrated that Cascade R-CNN++ outperforms Cascade R-CNN in terms of precision and detection quality. We conducted further analysis on multiresolution remote sensing images to verify model transferability across different object scales.
引用
收藏
页数:18
相关论文
共 45 条
  • [1] [Anonymous], 2017, P 2017 IEEE INT C CO
  • [2] [Anonymous], 2018, P IEEE C COMPUTER VI
  • [3] [Anonymous], 2016, Comput. Vis. Pattern Recogn.
  • [4] Bochkovskiy A., 2020, ARXIV 200410934
  • [5] Cai Z., P IEEE C 2018 COMPUT, P6154
  • [6] Cascade R-CNN: High Quality Object Detection and Instance Segmentation
    Cai, Zhaowei
    Vasconcelos, Nuno
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2021, 43 (05) : 1483 - 1498
  • [7] An automatic bridge detection technique for multispectral images
    Chaudhuri, D.
    Samal, Ashok
    [J]. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2008, 46 (09): : 2720 - 2727
  • [8] Guiding Clean Features for Object Detection in Remote Sensing Images
    Cheng, Gong
    He, Min
    Hong, Hailong
    Yao, Xiwen
    Qian, Xiaoliang
    Guo, Lei
    [J]. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 2022, 19
  • [9] Feature Enhancement Network for Object Detection in Optical Remote Sensing Images
    Cheng, Gong
    Lang, Chunbo
    Wu, Maoxiong
    Xie, Xingxing
    Yao, Xiwen
    Han, Junwei
    [J]. JOURNAL OF REMOTE SENSING, 2021, 2021
  • [10] Cross-Scale Feature Fusion for Object Detection in Optical Remote Sensing Images
    Cheng, Gong
    Si, Yongjie
    Hong, Hailong
    Yao, Xiwen
    Guo, Lei
    [J]. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 2021, 18 (03) : 431 - 435