针对巡飞器导航系统GPS拒止环境下无法定位及其航拍图像地物目标与背景混杂且目标边缘模糊导致分割精度低的问题，提出一种双源地理位置知觉地物目标实例分割模型（Dual source Geographic location aware Instance Segmentation Model，DGISM）。首先，设计沿水平、竖直和通道维度的三重自适应中心矩模块同时引入并行路径跨层交互结构聚合浅层细节信息，以提高实例特征区分度从而克服背景混杂问题。其次，设计多核空间聚焦模块融合跨尺度空间特征信息和深层通道信息以增强模型对边缘特征的聚集能力，改善了因边缘模糊导致分割精度低的问题。最后，针对GPS拒止环境下无法定位的问题，设计航拍图像与卫星信息双源感知定位模块，经分割权重提取巡飞器航拍图像特征信息与具有卫星定位信息的图像特征记忆库相匹配获取对应地理位置信息。实验结果表明，本文提出的双源地理位置知觉地物目标实例分割模型mAP50-95和mAP50分割精度分别达到69.1%和88.8%，图像特征双源匹配定位精度达到95.6%，在GPS拒止环境下分割建筑类地物目标实现巡飞器地理位置知觉精确定位。
To address the issues of inability to locate in GPS reject environments for the navigation system of aerial vehicles and the low segmentation accuracy caused by blurred edges and mixed backgrounds in aerial imagery, this paper proposes a dual-source geo-aware instance segmentation model for ground objects, referred to as DGISM. First, a triple adaptive central moment module along horizontal, vertical, and channel dimensions is designed, along with a parallel path cross-layer interaction structure to aggregate shallow-level details, thereby enhancing instance feature discriminability and overcoming background clutter issues. Second, a multi-core spatial focus module is developed to integrate cross-scale spatial features and deep channel information, improving the model's ability to concentrate on edge features and addressing the problem of low segmentation accuracy due to edge blur. Lastly, to tackle the inability to locate in GPS reject environments, a dual-source perception and positioning module for aerial imagery and satellite information is designed. This module extracts segmentation weights to match the feature information of the aerial vehicle's imagery with a feature memory bank of images containing satellite positioning information, thereby obtaining corresponding geographic location information. Experimental results demonstrate that the proposed dual-source geo-aware instance segmentation model achieves mAP50-95 and mAP50 segmentation accuracies of 69.1% and 88.8%, respectively, with a dual-source feature matching positioning accuracy of 95.6%. Segmentation of building-like terrestrial objects for aerial vehicle geographical location perception in GPS reject environments.
[1] 屈若锟, 王致远, 刘晔璐, 等．面向城市空中交通的无人机视觉定位技术研究[J]．航空学报, 2025, 46(10): 531168．
QU R K , WANG ZY, LIU Y L, et al. Research on UAV Visual Positioning Technology for Urban Air Mobility [J]. Acta Aeronautica et Astronautica Sinica, 2025,
46(10): 531168 (in Chinese)
[2] He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969.
[3] Chen K, Pang J, Wang J, et al. Hybrid task cascade for instance segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 4974-4983.
[4] Cheng B, Misra I, Schwing A G, et al. Masked-attention mask transformer for universal image segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022: 1290-1299.
[5] Chen X, Girshick R, He K, et al. Tensormask: A foundation for dense object segmentation[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 2061-2069.
[6] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.
[7] Wang X, Zhang R, Kong T, et al. Solov2: Dynamic and fast instance segmentation[J]. Advances in Neural information processing systems, 2020, 33: 17721-17732.
[8] Zhou C. Yolact++ better real-time instance segmentation[M]. University of California, Davis, 2020.
[9] Bolya D, Zhou C, Xiao F, et al. Yolact: Real-time instance segmentation[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 9157-9166.
[10] Dong C, Tang Y, Zhang L. GHA-Inst: a real-time instance segmentation model utilizing YOLO detection framework[J]. Cluster Computing, 2024, 27(6): 7401-7415.
[11] Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 7464-7475.
[12] 姚俞成，李旭，徐启敏，等．基于特征增强和校准的航拍车辆实例分割方法［J］．航空学报，2023， 44（24）：328397．
YAO Y C， LI X， XU Q M， et al. Instance segmentation for vehicle in UAV aerial images based on feature enhancement and calibration［J］. Acta Aeronautica et Astronautica Sinica， 2023， 44（24）： 328397 （in Chinese）
[13] Chen Y, Yuan X, Wang J, et al. YOLO-MS: rethinking multi-scale representation learning for real-time object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025.
[14] Wu H, Zhang M, Huang P, et al. CMLFormer: CNN and multi-scale local-context transformer network for remote sensing images semantic segmentation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024.
[15] Nanhua C, Liangyu Z. MMHCA: Multi-feature representations based on multi-scale hierarchical contextual aggregation for UAV-view geo-localization[J]. Chinese Journal of Aeronautics, 2024.
[16] 曹正阳,张冰,白屹轩,等.GNSS/INS/VNS组合定位信息融合的多无人机协同导航方法[J].兵工学报,2023,44(S2):157-166.
Caozhengyang, Zhang Bing, Bai Yixuan, et al. GNSS/INS/VNS integrated positioning information fusion method for multi UAV cooperative navigation [J]. Acta Armamentarii, 2023,44(S2):157-166 (in Chinese)
[17] Sun K, Mohta K, Pfrommer B, et al. Robust stereo visual inertial odometry for fast autonomous flight[J]. IEEE Robotics and Automation Letters, 2018, 3(2): 965-972.
[18] Nassar A, Amer K, ElHakim R, et al. A deep CNN-based framework for enhanced aerial imagery registration with applications to UAV geolocalization[C]//Proceedings of the IEEE conference on computer vision and pattern recognition workshops. 2018: 1513-1523.
[19] Choi J, Myung H. BRM localization: UAV localization in GNSS-denied environments based on matching of numerical map and UAV images[C]//2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020: 4537-4544.
[20] Ultralytics. YOLOv8 2023 [Available from: https://github.com/ultralytics/yolov8.
[21] 邓立暖,尧新峰. 基于NSST的红外与可见光图像融合算法[J]. 电子学报,2017,45(12):2965-2970.
Denglinuan, Yao Xinfeng, et al. Infrared and visible image fusion algorithm based on NSST [J]. Journal of Electronics, 2017,45(12):2965-2970 (in Chinese)
[22] Dai M, Zheng E, Feng Z, et al. Vision-based UAV self-positioning in low-altitude urban environments[J]. IEEE Transactions on Image Processing, 2023, 33: 493-508.
[23] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141.
[24] Roy A G, Navab N, Wachinger C. Concurrent spatial and channel ‘squeeze & excitation’in fully convolutional networks[C]//Medical Image Computing and Computer Assisted Intervention–MICCAI 2018: 21st International Conference, Granada, Spain, September 16-20, 2018, Proceedings, Part I. Springer International Publishing, 2018: 421-429.
[25] Cao Y, Xu J, Lin S, et al. Gcnet: Non-local networks meet squeeze-excitation networks and beyond[C]//Proceedings of the IEEE/CVF international conference on computer vision workshops. 2019: 0-0.
[26] Zhang P, Ran H, Jia C, et al. A lightweight propagation path aggregating network with neural topic model for rumor detection[J]. Neurocomputing, 2021, 458: 468-477.
[27] Zhang X, Zeng H, Guo S, et al. Efficient long-range attention network for image super-resolution[C]//European conference on computer vision. Cham: Springer Nature Switzerland, 2022: 649-667.
CJA