基于图像间的语义感知协同目标检测算法  

Cooperative Salient Object Detection Algorithm Based on Semantic Perception between Images

作　　者：葛延良[1] 张静[1] 毕洪波[1] 张林 田鸿鹏 贺敏 李德鑫 GE Yan-liang;ZHANG Jing;BI Hong-bo;ZHANG Lin;TIAN Hong-peng;HE Min;LI De-xin(School of Electrical&Information Engineering,Northeast Petroleum University)

机构地区：[1]东北石油大学电气信息工程学院

出　　处：《化工自动化及仪表》2024年第1期56-62,共7页Control and Instruments in Chemical Industry

基　　金：黑龙江省自然科学基金(批准号:LH2022F005)资助的课题。

摘　　要：提出了一个基于图像间的语义感知协同目标检测算法,以高效的方式检索整个图像组的共识线索。对每个像素用语义感知协同模块(SPCM)获取其交叉路径上所有像素的上下文信息。通过进一步的操作,最终可以捕获整个图像的共识信息。辅助分类融合模块(ACFM)可使网络以自上而下的方式突出协同区域。协同显著性检测实验结果表明,文中的算法在3种常用的iCoseg、Cosal2015和CoSOD3k数据集中表现良好,4个指标性能优异,验证了该方法的有效性。In this paper,a cooperative salient object detection algorithm based on semantic perception between images was proposed,which can retrieve consensus clues of the entire image group in an efficient way.As for each pixel,a novel semantic perception collaborative module(SPCM)was used to obtain the context information of all pixels on cross path.Through further operation,each pixel can finally capture consensus information of the entire image.In addition,a new auxiliary classification fusion module(ACFM)was proposed to make network highlight the collaborative region in a top-down manner.The experimental results show that,this algorithm works well in commonly-used iCoseg,Cosal2015 and CoSOD3k data sets,the four indexes of performance are excellent to further verify the effectiveness of this method.

关 键 词：协同显著性检测 图像识别 语义感知协同模块 辅助分类融合模块 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]