兼顾语义与图像特征的外版海图底质注记高效解译算法

doi:10.11947/j.AGCS.2026.20260115

摘要/Abstract

摘要：

针对当前外版栅格海图底质注记自动解译中存在的复杂场景适应性较差、效率较低及语义解译割裂等问题，本文提出一种兼顾语义与图像特征的外版海图底质注记高效解译算法。首先，构建基于底质注记图像表达特征的图像优化策略，通过陆地区域掩膜与自适应阈值分割，有效滤除栅格海图复杂背景干扰。然后，引入模板匹配技术，建立基于Hu不变矩特征的底质注记模板库，实现字符识别与语义解析的一体化解译。最后，针对复合注记模板构建困难问题，设计顾及语义完整性的两级注记切片分割模型，解决了复杂组合注记的解译难题。试验结果表明：①本文算法成功实现了外版海图底质注记的高效解译，在底质注记识别与语义解析的过程中，具有较高的分类查准率、查全率、识别完整率及语义准确。②通过与现有算法在不同复杂度场景下的对比试验表明，本文算法在保持更高识别精度的同时，进一步提高了整体处理效率，且在中、高复杂度场景中有效解决了对比算法存在的注记漏识别与语义解析错误问题，证明本文算法具有更高的稳健性与效率。

关键词: 海图信息化, 自动提取, 底质注记, 语义解析, 模板匹配

Abstract:

In response to current issues in the automatic interpretation of foreign raster nautical chart substrate annotations, such as poor adaptability to complex scenarios, low efficiency, and the disconnection between semantics interpretation and image processing, this paper proposes an efficient interpretation algorithm for foreign nautical chart substrate annotations considering semantics and image features. Firstly, the algorithm constructs an image optimization strategy based on the visual representation features of substrate annotations. By applying land area masking and adaptive threshold segmentation, it effectively filters out interference from the complex background of raster charts. Subsequently, template matching technology is introduced to establish a substrate annotation template library based on Hu invariant moment features, achieving integrated interpretation for both character recognition and semantic parsing. Finally, to address the difficulty in constructing templates for composite annotations, a two-level annotation slice segmentation model considering semantic integrity is designed, solving the interpretation challenge of complex combined annotations. Experimental results show that: ①The proposed algorithm successfully achieves efficient interpretation of foreign chart substrate annotations, demonstrating high precision, recall, recognition completeness rate, and semantic accuracy during the processes of annotation recognition and semantic parsing.②Comparative experiments with existing methods under scenarios of varying complexity indicate that the proposed algorithm maintains higher recognition accuracy while further improving overall processing efficiency. Moreover, it effectively addresses issues of missed annotations and semantic parsing errors encountered by comparative methods in medium-and high-complexity scenarios, proving the superior robustness and efficiency of our approach.

Key words: nautical chart informatization, automatic extraction, substrate annotations, semantic parsing, template matching

中图分类号:

P285

董箭, 谭超宇, 唐露露, 戴泽源, 谢天, 陈长林, 倪尔. 兼顾语义与图像特征的外版海图底质注记高效解译算法[J]. 测绘学报, 2026, 55(9): 1567-1581.

Jian Dong, Chaoyu Tan, Lulu Tang, Zeyuan Dai, Tian Xie, Changlin Chen, Er Ni. Efficient interpretation algorithm for foreign nautical chart substrate annotations considering semantics and image features[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(9): 1567-1581.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://xb.chinasmp.com/CN/10.11947/j.AGCS.2026.20260115

http://xb.chinasmp.com/CN/Y2026/V55/I9/1567

图/表 20

图1

图2

图3

图4

图5

表1

表2

图6

表3

表4

表5

表6

表7

图7

表8

图8

表9

图9

表10

图10

参考文献 31

[1]	阳天舒, 李梅, 信荟敏, 等. 基于形态学的自适应阈值分割算法[J]. 电子设计工程, 2015, 23(13): 102-104.
	Yang Tianshu, Li Mei, Xin Huimin, et al. The adaptive threshold segmentation algorithm based on morphology[J]. Electronic Design Engineering, 2015, 23(13): 102-104.
[2]	Yan F X, Zhang H, Kube C R. A multistage adaptive thresholding method[J]. Pattern Recognition Letters, 2005, 26(8): 1183-1191.
[3]	陈亚军, 刘丁, 梁军利, 等. 一种无须预指定分割区域数的自适应多阈值图像分割方法[J]. 模式识别与人工智能, 2014, 27(11): 993-1004.
	Chen Yajun, Liu Ding, Liang Junli, et al. An adaptive multi-threshold image segmentation method without preassigning the number of segmented regions[J]. Pattern Recognition and Artificial Intelligence, 2014, 27(11): 993-1004.
[4]	龙建武, 李继豪, 曾谁飞. 自适应多阈值图像分割算法[J]. 通信学报, 2025, 46(8): 241-255.
	Long Jianwu, Li Jihao, Zeng Shuifei. Adaptive multi-threshold image segmentation algorithm[J]. Journal on Communications, 2025, 46(8): 241-255.
[5]	Bohat V K, Arya K V. A new heuristic for multilevel thresholding of images[J]. Expert Systems with Applications, 2019, 117: 176-203.
[6]	Ojala T, Pietikainen M, Maenpaa T. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(7): 971-987.
[7]	Alhindi T J, Kalra S, Ng K H, et al. Comparing LBP, HOG and deep features for classification of histopathology images[C]//Proceedings of 2018 International Joint Conference on Neural Networks. Rio de Janeiro: IEEE, 2018: 5384-5390.
[8]	刘启浪, 汤勃, 孔建益, 等. 基于多尺度LBP特征的带钢表面缺陷图像SVM分类[J]. 组合机床与自动化加工技术, 2020(12): 27-30.
	Liu Qilang, Tang Bo, Kong Jianyi, et al. SVM classification of surface defect images of strip based on multi-scale LBP features[J]. Modular Machine Tool & Automatic Manufacturing Technique, 2020(12): 27-30.
[9]	吴啸天. 基于多特征提取和SVM分类器的木材显微识别[J]. 陕西林业科技, 2017, 45(5): 11-14.
	Wu Xiaotian. Wood microscopic recognition based on feature extraction and SVM classifiers[J]. Shaanxi Forest Science and Technology, 2017, 45(5): 11-14.
[10]	Al Saidi I, Rziza M, Debayle J. Completed homogeneous LBP for remote sensing image classification[J]. International Journal of Remote Sensing, 2023, 44(12): 3815-3836.
[11]	Veerashetty S, Patil N B. Novel LBP based texture descriptor for rotation, illumination and scale invariance for image texture analysis and classification using multi-kernel SVM[J]. Multimedia Tools and Applications, 2020, 79(15): 9935-9955.
[12]	Younesi A, Ansari M, Fazli M, et al. A comprehensive survey of convolutions in deep learning: applications, challenges, and future trends[J]. IEEE Access, 2024, 12: 41180-41218.
[13]	Gao Jinfeng, Chen Yu, Wei Yongming, et al. Detection of specific building in remote sensing images using a novel YOLO-S-CIOU model. case: gas station identification[J]. Sensors, 2021, 21(4): 1375.
[14]	姚宇捷, 彭育辉, 陈泽辉, 等. 支持抗光照目标检测的改进YOLO算法[J]. 汽车工程, 2023, 45(5): 777-785.
	Yao Yujie, Peng Yuhui, Chen Zehui, et al. An improved YOLO algorithm supporting anti-illumination target detection[J]. Automotive Engineering, 2023, 45(5): 777-785.
[15]	孙培双, 温显斌. 基于改进YOLOv5模型的SAR图像舰船目标检测算法[J]. 电光与控制, 2024, 31(8): 32-37, 85.
	Sun Peishuang, Wen Xianbin. An improved algorithm for detecting ship target in SAR images based on YOLOv5 model[J]. Electronics Optics & Control, 2024, 31(8): 32-37, 85.
[16]	田丽, 贾煜辉. 改进YOLOv5s算法的高光谱遥感图像目标检测[J]. 吉林大学学报(工学版), 2025, 55(5): 1742-1748.
	Tian Li, Jia Yuhui. Improved YOLOv5s algorithm for target detection in hyperspectral remote sensing images[J]. Journal of Jilin University (Engineering and Technology Edition), 2025, 55(5): 1742-1748.
[17]	禄小敏, 张志义, 闫浩文, 等. 融合深度图信息最大化和多层感知机的建筑物群组模式识别方法[J]. 测绘学报, 2026, 55(3): 425-438. DOI: . doi: 10.11947/j.AGCS.2026.20250348
	Lu Xiaomin, Zhang Zhiyi, Yan Haowen, et al. A recognition method for building group pattern integrating deep graph infomax and multilayer perceptron[J]. Acta Geodaetica et Cartographica Sinica, 2026, 55(3): 425-438. DOI: . doi: 10.11947/j.AGCS.2026.20250348
[18]	Shi Baoguang, Bai Xiang, Yao Cong. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(11): 2298-2304.
[19]	舒珊珊, 郑晓旭, 文成玉. 基于CRNN改进的中文手写体文本行识别[J]. 成都信息工程大学学报, 2023, 38(4): 422-428.
	Shu Shanshan, Zheng Xiaoxu, Wen Chengyu. Improved Chinese handwritten text line recognition based on CRNN[J]. Journal of Chengdu University of Information Technology, 2023, 38(4): 422-428.
[20]	曹叶欢. 基于CRNN的质量文档识别系统[J]. 信息技术与信息化, 2024(10): 53-56.
	Cao Yehuan. Quality document recognition system based on CRNN[J]. Information Technology and Informatization. 2024(10): 53-56.
[21]	Trinh T D, Le T A D, Nguyen N T, et al. An improved CRNN for Vietnamese identity card information recognition[J]. Computer Systems Science and Engineering, 2022, 40(2): 539-555.
[22]	Ma Mengkai, Dong Jian, Tang Lulu, et al. Semantic-aware automatic extraction method for bottom sediment annotations in raster nautical charts[J]. Cartography and Geographic Information Science, 2024, 51(6): 744-760.
[23]	Annaby M, Fouda Y. Fast template matching and object detection techniques usingϕ-correlation and binary circuits[J]. Multimedia Tools and Applications, 2024, 83(3): 6469-6496.
[24]	Dong Jing, Hu Maoqing, Lu Jiazhen, et al. Affine template matching based on multi-scale dense structure principal direction[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2021, 31(6): 2125-2132.
[25]	段亮弟, 宋平, 陈众, 等. 基于均值漂移的快速模板匹配算法[J]. 西北工业大学学报, 2018, 36(4): 792-799.
	Duan Liangdi, Song Ping, Chen Zhong, et al. A target tracking algorithm based on mean shift and fast template matching[J]. Journal of Northwestern Polytechnical University, 2018, 36(4): 792-799.
[26]	袁狄剑, 许新科, 刘通, 等. 基于自构卷积核改进模板匹配的红外十字靶标中心定位算法[J]. 红外与毫米波学报, 2025, 44(3): 469-476.
	Yuan Dijian, Xu Xinke, Liu Tong, et al. An improved template matching algorithm for infrared cross-target center positioning based on self-constructed convolution kernels[J]. Journal of Infrared and Millimeter Waves, 2025, 44(3): 469-476.
[27]	张伟, 何金国. Hu不变矩的构造与推广[J]. 计算机应用, 2010, 30(9): 2449-2452.
	Zhang Wei, He Jinguo. Construction and generalization of Hu moment invariants[J]. Journal of Computer Applications, 2010, 30(9): 2449-2452.
[28]	Liberti L, Lavor C, Maculan N, et al. Euclidean distance geometry and applications[J]. SIAM Review, 2014, 56(1): 3-69.
[29]	Yun S, Han D, Chun S, et al. CutMix: regularization strategy to train strong classifiers with localizable features[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 6022-6031.
[30]	Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 936-944.
[31]	Shi Baoguang, Bai Xiang, Yao Cong. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(11): 2298-2304.

序号	Id号	原图底质	是否被识别	注记识别是否完整	语义信息提取是否精确
1	1	fS	是	是	精确提取：fine sand
2	2	fS. bkSh	是	是	精确提取：fine sand broken shells
3	3	S. G. bkSh	是	是	精确提取：sand gravel broken shells
4	4	S. bkSh	是	是	精确提取：sand broken shells
5	5	soM. fS. bkSh	是	是	精确提取：soft mud fine sand broken shells
⋮	⋮	⋮	⋮	⋮	⋮
35	35	M. S	是	否	语义缺失：sand
78	78	soM	是	否	语义缺失：soft
114	114	S. G. bkS	是	否	语义缺失：sand gravel broken sand

轮次	迭代数	起始学习率	学习率变化方式	批次大小
冻结轮次	50	10^-3	阶梯式衰减 lr=lr×α（α=0.9）不变化	4
解冻轮次	50	10^-4	阶梯式衰减 lr=lr×α（α=0.9）不变化	4

注记种类	查准率/（%）	查全率/（%）	语义提取准确率/（%）	处理时间/s
本文模型	100	100	100	5.95+0.28 （识别+语义解译）
简化模型A	100	100	100	5.95+19.65 （识别+语义解译）
简化模型B	100	91.01	100	8.32+0.28 （识别+语义解译）
文献[22]模型	100	91.01	100	8.32+22.38 （识别+语义解译）
现有模型（CRNN）	100	91.01	100	8.32+17.21 （识别+语义解译）

注记种类	查准率/（%）	查全率/（%）	语义提取准确率/（%）	处理时间/s
本文模型	100	91.01	100	6.11+0.25 （识别+语义解译）
简化模型A	100	91.01	100	6.11+20.87 （识别+语义解译）
简化模型B	100	81.82	88.89	9.06+0.25 （识别+语义解译）
文献[22]模型	100	81.82	88.89	9.06+24.26 （识别+语义解译）
现有模型（CRNN）	100	81.82	100	9.06+17.64 （识别+语义解译）