舰船种类丰富、大小差异显著，从小型快艇到大型航空母舰，尺度范围跨度大，给舰船目标类型识别检测带来较大难度。为解决这一问题，提出了基于动量-自适应人工神经网络的舰船目标检测方法。此方法利用最大类间方差算法，结合图像灰度特性，通过计算类间方差最大化，将图像划分为目标与背景；采用最小外接矩形法提取舰船目标特征，通过计算能够完全包围舰船的最小矩形，获取目标的几何特征信息；将所提取特征输入基于改进人工神经网络的舰船目标检测模型，模型在权重调节中引入动量项，设计自适应学习率，增强网络对舰船特征的学习与识别能力，实现对舰船目标的识别检测。实验结果验证，该方法能够清晰地将舰船目标从背景中分离出来；对于单个舰船目标类型以及多目标类型的识别检测结果均稳定、准确。
There are a wide variety of ship types with significant differences in size, ranging from small speedboats to large aircraft carriers, which pose great difficulties in identifying and detecting ship target types. To address this issue, a ship target detection method based on momentum adaptive artificial neural network is proposed. This method utilizes the maximum inter class variance algorithm, combined with the grayscale characteristics of the image, to divide the image into target and background by calculating the maximum inter class variance; Using the minimum bounding rectangle method to extract ship target features, the geometric feature information of the target is obtained by calculating the minimum rectangle that can completely surround the ship; Input the extracted features into a ship target detection model based on an improved artificial neural network. The model introduces momentum terms in weight adjustment, designs an adaptive learning rate, enhances the network's ability to learn and recognize ship features, and achieves recognition and detection of ship targets. Experimental results verification: This method can clearly separate ship targets from the background; The recognition and detection results for both single ship target types and multiple target types are stable and accurate.
2025,47(11): 175-179 收稿日期：2025-1-12
DOI：10.3404/j.issn.1672-7649.2025.11.031
分类号：TP391
作者简介：贾世杰(2002-),男,硕士,研究方向为计算机科学及人工智能
参考文献：
[1] 赵其昌, 吴一全, 苑玉彬. 光学遥感图像舰船目标检测与识别方法研究进展[J]. 航空学报, 2024, 45(8): 46-79.
[2] 许京新, 王金伟, 宋富骏, 等. 融合自适应卷积神经网络的SAR图像舰船目标检测方法[J]. 信号处理, 2024, 40(9): 1696-1708.
[3] 杨明秋, 左小清, 董燕. 基于改进YOLOv8s的SAR图像舰船目标检测方法[J]. 激光与光电子学进展, 2024, 61(22): 296-309.
[4] 扈琪, 胡绍海, 刘帅奇. 基于多层显著性模型的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2024, 46(2): 478-487.
[5] 张炳焱, 张闯, 石振男, 等. 基于YOLO-FNC模型的轻量化船舶检测方法[J]. 中国舰船研究, 2024, 19(5): 180-187.
[6] 徐继尚, 柳翠寅, 刘明. 改进YOLOV4-Tiny的SAR图像舰船小目标检测[J]. 小型微型计算机系统, 2023, 44(12): 2791-2796.
[7] 陈玺, 何斌, 龙勇机, 等. 复杂海背景下的自适应舰船目标检测[J]. 液晶与显示, 2022, 37(3): 405-414.
