同步码字优化降噪的声纳图像多目标检测方法  

Multi-object detection method of sonar image with simultaneous codeword optimization for noise reduction

作　　者：魏光春 邢传玺[1] 崔晶 董赛蒙 WEI Guang chun;XING Chuan xi;CUI Jing;DONG Sai meng(School of Electrical and Information Technology,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学电气信息工程学院,云南昆明650500

出　　处：《海洋测绘》2024年第3期42-46,共5页Hydrographic Surveying and Charting

基　　金：国家自然科学基金(61761048);云南省基础研究专项面上项目(202101AT070132)。

摘　　要：针对海底侧扫声纳图像分辨率低、噪声污染严重导致水下目标检测不准确的问题,提出一种结合同步码字优化降噪的水下声纳图像目标检测方法。利用同步码字优化对声纳图像中的乘性噪声进行降噪处理,从而使图像中的水下目标物获得更好的视觉与检测效果,同时对声纳图像进行相应的数据集扩充。最后利用适合本文方法的YOLO系列中的YOLOv7对降噪后声纳图像中的目标物体进行检测,并在其特征网络中加入了卷积块注意模块,从而加强对目标的特征提取。仿真结果分析得出,同步码字优化降噪与YOLOv7相结合的目标检测方法,可使目标置信度达到79%,相较于降噪前的目标检测置信度提高16%,对于目标较小的物体,能更好地改善漏检与误检情况。In order to solve the problem of inaccurate identification of underwater targets due to low resolution and serious noise pollution in undersea side-scan sonar images,this paper proposes a target detection method for underwater sonar images combined with simultaneous codeword optimization for noise reduction.The multiplicative noise in the sonar image is reduced by using simultaneous code word optimization,so that the underwater targets in the image can be better visualized and detected,and the corresponding data set is expanded for the sonar image.Finally,YOLOv7 in the YOLO series,which is suitable for the method in this paper,is used to detect the target objects in the sonar images after noise reduction,and the convolutional block attention module is added to its feature network to enhance the feature extraction of the targets.The analysis of simulation results concludes that the target detection method combining synchronous codeword optimized noise reduction and YOLOv7 can achieve a target confidence of 78%,which is 15%higher compared with the target detection confidence before noise reduction,and can better improve the leakage and false detection for objects with smaller targets.

关 键 词：侧扫声纳图像处理 水下目标特征提取 多目标检测 同步码字优化降噪 YOLOv7目标识别 

分 类 号：P229[天文地球—大地测量学与测量工程]