扫 描 看 全 文
引用
张泽琨，谭震彪，余坤，方斌，黄骁，马杰.XXXX.特征图强化网络：一种利用特征图强化船舶检测模型训练的网络结构.遥感学报，XX（XX）： 1-9
Zhang Zekun，Tan Zhenbiao，Yu Kun，Fang Bing，Huang Xiao，Ma Jie. XXXX. FMRNet： A network structure for enhanced ship detection model training using feature maps. National Remote Sensing Bulletin， XX（XX）：1-9
随着人工智能技术的进一步发展，深度学习方法在船舶检测领域发挥着重要作用。然而，深度学习算法出现的虚警和漏检，对船舶检测领域技术的应用存在一定的阻碍。虽然经典的深度学习方法能够有效处理单一背景的海面，但是当面对复杂背景之下的数据时，经典模型很容易得出岸上的虚警。并且在常规训练中，模型常常对一些显著特征过于关注，出现特征过拟合现象，当这些显著特征发生改变时极易出现漏检。在模型对输入进行前向传播的过程中，模型中不同网络层会对输入生成对应的映射，也就是特征图。充分利用特征图的语义信息和空间信息是一种有效减少虚警和漏检的方法。与传统模型相比，我们提出的特征图强化网络可以充分利用特征图生成自适应特征图掩码与水陆分割掩码，避免模型的特征过拟合与削弱复杂背景造成的影响，最终达到减少虚警与漏检的目的。在现有公共数据集上与算法模型的对比实验结果表明，本文所提出的方法的性能更为出色，超过了其他SOTA算法。
With the further development of artificial intelligence technology, deep learning methods play an important role in the field of ship detection. However, the false alarms and missed detections that appear in deep learning algorithms are a hindrance to the application of technology in the field of ship detection. Although the classical deep learning methods can effectively deal with a single background sea surface, when faced with data under complex backgrounds, the classical models can easily yield false alarms on shore. And in custom training, the model often pays too much attention to some salient features that feature overfitting occurs. It is extremely easy to miss detection when these salient features change. In the process of forward propagation of the model to the input, different network layers in the model generate corresponding mappings or feature maps from the input. Making full use of the semantic and spatial information of the feature maps is an effective way to reduce false alarms and missed detections. Compared with the traditional model, our proposed Feature Map Reinforcement Network (FMRNnet) can make full use of feature maps to generate adaptive feature map masks and water-land segmentation masks. It ultimately achieves the goal of reducing false alarms and missed detections by avoiding feature overfitting of the model and weakening the effects caused by complex backgrounds. In FMRNnet, we design the Self Feature-map Mask Module (SFMM), which can selectively utilize the feature map through the attention mechanism to generate an adaptive mask. The mask prevents the model from focusing on a single feature point, thus preventing feature overfitting. We also propose a Feature-map Sea-land Segmentation Module (FSSM) that is parallel to SFMM. It reduces the false alarms of ship targets appearing in the land area by introducing the fusion between the water-land segmentation mask and the feature map. The experimental results compared with SOTA algorithms on the existing public dataset show that the performance of the proposed method in this paper is excellent and outperforms other SOTA algorithms. After adding FMRNet, the ten-fold average mAP value of the detection algorithm ROI trans has a large improvement. This improves the mean value of baseline mAP from 86.1% to 90.8% and surpasses other SOTA. Benefiting from the adaptive mask, the mAP value of the model including the SFMM module is 90.4%, achieving a 4.2% improvement over the baseline. Due to the priori knowledge learned from the water-land distribution, FSSM improves the precision and recall of the model and achieves a MAP value of 86.4%. For the task of ship detection, we propose a novel backbone network, FMRNet based on Resnet. To avoid overfitting salient features, our proposed SFMM module enables the model to judge the target from multiple features. To reduce the false alarms caused by complex backgrounds, we designed the FSSM module. By suppressing the non-water surface area, it reduces the confidence level of targets appearing in non-water surface. FSSM achieves the purpose of removing unreasonable false alarms while improving the accuracy of the model.
遥感成像人工智能船舶目标检测神经网络特征提取
Remote sensingArtificial IntelligenceShip Object DetectionNeural NetworkFeature map Extraction
Bi F, Hou J, Chen L, Yang Z, and Wang Y. 2019. Ship detection for optical remote sensing images based on visual attention enhanced network. Sensors, 2271 [DOI:19(10), 2271 10.3390/s19102271http://dx.doi.org/10.3390/s19102271]
Ding G, Kha S, and Tang Z. 2020. Let Features Decide for Themselves: Feature Mask Network for Person Re-identification. Pattern Recognition Letters, 91-98. arXiv:1711.07155
Ding J, Xue N, and Xia G. 2018. Learning RoI Transformer for Detecting Oriented Objects in Aerial Images. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1 Dec 2018, arXiv: 1812.00155
Girshick R. 2015. Fast R-CNN. International Conference on Computer Vision (ICCV), Santiago: 1440-1448 [DOI: 10.1109/ICCV.2015.169http://dx.doi.org/10.1109/ICCV.2015.169]
Goetz A, Rock B N, and Rowan L C. 1983. Remote sensing for exploration - An overview, 573-590 [DOI:10.1190/1.1438015http://dx.doi.org/10.1190/1.1438015]
Gu Z. 2010. Research on high performance computing platform of ocean environmental information processing. 2010 International Conference on Optics, Photonics and Energy Engineering (OPEE), 298-300 [DOI: 10.1109/OPEE.2010.5508128http://dx.doi.org/10.1109/OPEE.2010.5508128]
He K, Zhang X, Ren S, and Sun J. 2016. Deep Residual Learning for Image Recognition. in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), vol. 4: 770–778 [DOI: 10.1109/CVPR.2016.90http://dx.doi.org/10.1109/CVPR.2016.90]
Hu J, Shen L, and Sun G. 2018. Squeeze-and excitation networks. in Proc. of IEEE Conference on Computer Vision and Pattern Recognition (CVPR): 7132–7141 [DOI: 10.1109/CVPR.2018.00745http://dx.doi.org/10.1109/CVPR.2018.00745]
Liu Z,Yuan L,and Weng L. 2017. A High Resolution Optical Satellite Image Dataset for Ship Recognition and Some New Baselines. ICPRAM. 324-331
Lunetta R, Congalton R, Fenstermaker L, Jensen J, and Tinney L. 1991. Remote-sensing and geographic information-system data integration - error sources and research issues. Photogrammetric Engineering & Remote Sensing, 57(6): 677-687 [DOI: 10.1016/0031-0182(91)90167-Phttp://dx.doi.org/10.1016/0031-0182(91)90167-P]
Nie X, Duan M, Ding H, Hu B and Wong B. 2020. Attention mask r-cnn for ship detection and segmentation from remote sensing images. IEEE Access, 2020, 8, 9325-9334 [DOI:10.1109/ACCESS.2020.2964540http://dx.doi.org/10.1109/ACCESS.2020.2964540]
Qian W, Yang W, and Peng S. 2019. Learning modulated loss for rotated object detection. arXiv:1911.08299
Redmon J, Divvala S, Girshick R and Farhadi A. 2016. You only look once: unified, real-time object detection. Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition, 779- 788 [DOI: 10.1109/CVPR.2016.91http://dx.doi.org/10.1109/CVPR.2016.91]
Redmon J, and Farhadi A. 2017. YOLO9000: better, faster, stronger. Proceedings of the IEEE conference on computer vision and pattern recognition, 7263-7271 [DOI: 10.1109/CVPR.2017.690http://dx.doi.org/10.1109/CVPR.2017.690]
Redmon J and Farhadi A. 2018. YOLOv3: An incremental improvement. arXiv:1804.02767v1
Simonyan K, and Zisserman A. 2015. Deep Convolutional Networks for Large-Scale Image Recognition. ICLR, 1-14.
许开宇，胡文骅，陈仁一种基于Canny算子的船舶目标检测算法. 2006. 《红外技术》2006年 第6期. 315-318页. [DOI: 10.3969/j.issn.1001-8891.2006.06.002http://dx.doi.org/10.3969/j.issn.1001-8891.2006.06.002]
Xu Y, Fu M, Wang Qand X. 2021. Gliding. Vertex on the Horizontal Bounding Box for Multi-Oriented Object Detection. IEEE Trans. Pattern Anal. Mach. Intell. 2021, 43(3), 1452-1459 [DOI:10.1109/TPAMI.2020.2974745http://dx.doi.org/10.1109/TPAMI.2020.2974745]
Yang X，Chi Y，Zhou Y and Wang Y. 2021. Spectral-Spatial Attention Bilateral Network for Hyperspectral Image Classification. Journal of Remote Sensing. 0-0
杨星，池越，周亚同，王杨. 2021. 基于光谱-空间注意力双边网络的高光谱图像分类. 遥感学报, 0-0[DOI: 10.11834/jrs.20210563http://dx.doi.org/10.11834/jrs.20210563]
Yang X, Liu Q, and Yan J. R3Det: Refined single-stage detector with feature refinement for rotating object. arXiv: 1908.05612.
相关作者
相关机构
京公网安备11010802024621
