复杂场景下的SAR图像船舶目标检测

Ship target detection of SAR images in complex scenes

针对SAR图像中复杂场景下的船舶目标检测问题,提出一种将定位、分类和分割多任务结合一起的Mask-FPN模型,在金字塔特征映射图的基础上,同时引入图像分割分支,并通过多任务损失函数保证定位分类和分割三个过程同时进行.实验结果表明,本文提出的Mask-FNP模型在近海、港口、岛屿等存在干扰背景的复杂场景下,船舶识别准确率达98.81%.与Faster R-CNN、Yolo3、SSD、FPN等模型对比,本文模型检测准确率提高,虚警率和漏检率明显下降.

Aiming at the problem of ship target detection in complex scene of SAR image,a Mask-FPN model combining localization,classification and segmentation multitasking was proposed.Based on pyramid feature map,the image segmentation branches were introduced at the same time,and multi task loss function was used to ensure that the three processes as positioning,classification and segmentation were carried out simultaneously.The experimental results show that the accuracy of the Mask-FNP model is up to 98.81%in the complex scenes with interference background such as offshore,port and island.Compared with Faster R-CNN,Yolo3,SSD,FPN and other models,the detection accuracy of this model is higher,and the false alarm rate and missing detection rate are significantly reduced.