在复杂的海面环境中，因不规律变换的海面浪花、岸边岛屿以及舰船阴影等因素影响，使海面舰船目标检测成为一个难点，为此，提出基于高斯混合模型的海面舰船目标检测方法。结合多尺度中值滤波和Canny边缘检测处理采集海面舰船图像并获取边缘图像，采用霍夫变换从边缘图像中提取候选直线，并综合长度特征与颜色特征，从候选直线里精确提取海天线，完成舰船图像背景补偿。以背景补偿后图像为基础，建立高斯混合模型，通过模型初始化、修正以及背景衡量和前景划分等步骤检测舰船目标，并结合霍特林正交变换抑制舰船目标阴影，完成图像后处理。实验结果说明，中值滤波的尺度为2时图像清晰度最好。该方法可有效提取海天线，完成舰船图像背景补偿，精准检测海面舰船目标，结合阴影抑制能提高舰船目标检测精度，满足海面舰船目标检测的需要。
In complex sea environments, the detection of ship targets on the sea surface becomes a technological challenge due to factors such as irregular changes in sea waves, shore islands, and ship shadows. To address this issue, a Gaussian mixture model based method for ship target detection on the sea surface is proposed. Combining multi-scale median filtering and Canny edge detection processing to collect sea surface ship images and obtain edge images, the Hough transform is used to extract candidate lines from the edge image, and the length and color features are integrated to accurately extract sea antennas from the candidate lines. The background compensation of the ship image is completed, and a Gaussian mixture model is established based on the background compensated image. Through model initialization Correction, background measurement, and foreground segmentation are used to detect ship targets, and combined with Hotelling orthogonal transformation to suppress ship target shadows, completing image post-processing. The experimental results show that the image clarity is best when the scale of median filtering is 2. This method can effectively extract sea antennas, complete background compensation for ship images, and accurately detect ship targets on the sea. Combining shadow suppression can improve the accuracy of ship target detection and meet the needs of ship target detection on the sea.
2024,46(1): 148-151 收稿日期：2023-06-23
DOI：10.3404/j.issn.1672-7649.2024.01.025
分类号：TP391
作者简介：向文豪(1981-),男,博士,研究员,研究方向为舰载无人机任务系统
参考文献：
[1] 杨雨涵, 惠斌, 常铮, 等. 基于高斯混合模型的海上运动目标检测算法[J]. 计算机应用研究, 2020, 37(S1): 310-313.
[2] 刘伟, 郝晓丽, 吕进来. 自适应混合高斯建模的高效运动目标检测[J]. 中国图象图形学报, 2020, 25(1): 113-125.
[3] 刘江, 郭荣春, 王燕妮. 基于卡尔曼滤波的高斯混合模型目标检测算法[J]. 探测与控制学报, 2022, 44(2): 79-84.
[4] 汤旻安, 王晨雨. 基于改进ViBe算法的静态场景运动目标检测[J]. 激光与光电子学进展, 2021, 58(14): 216-224.
[5] 李阳, 华驰, 唐庭龙. 基于RPCA和SVM的快速运动目标检测算法[J]. 计算机仿真, 2022, 39(11): 463-466+495.
[6] 房玉杰, 张松, 刘晋, 等. 基于高斯混合模型的脚步声身份识别方法[J]. 计算机应用, 2021, 41(S2): 221-225.
