基于多特征融合卷积神经网络的显著性检测  

Saliency detection based on multi-feature fusion convolutional neural network

作　　者：赵应丁[1] 岳星宇 杨文姬[1,4] 张吉昊 杨红云 ZHAO Ying-ding;YUE Xing-yu;YANG Wen-ji;ZHANG Ji-hao;YANG Hong-yun(School of Software,Jiangxi Agricultural University,Nanchang 330045;School of Computer and Information Engineering,Jiangxi Agricultural University,Nanchang 330045;School of Foreign Languages,Huazhong University of Science and Technology,Wuhan 430074;Key Laboratory of Agricultural Information Technology of Colleges and Universities in Jiangxi Province,Nanchang 330045,China)

机构地区：[1]江西农业大学软件学院,江西南昌330045 [2]江西农业大学计算机与信息工程学院,江西南昌330045 [3]华中科技大学外国语学院,湖北武汉430074 [4]江西省高等学校农业信息技术重点实验室,江西南昌330045

出　　处：《计算机工程与科学》2021年第4期729-737,共9页Computer Engineering & Science

基　　金：国家自然科学基金(61462038,61562039);江西省教育厅科技计划项目(GJJ190217)。

摘　　要：随着深度学习技术的发展以及卷积神经网络在众多计算机视觉任务中的突出表现,基于卷积神经网络的深度显著性检测方法成为显著性检测领域的主流方法。但是,卷积神经网络受卷积核尺寸的限制,在网络底层只能在较小范围内提取特征,不能很好地检测区域内不显著但全局显著的对象;其次,卷积神经网络通过堆叠卷积层的方式可获得图像的全局信息,但在信息由浅向深传递时,会导致信息遗失,同时堆叠太深也会导致网络难以优化。基于此,提出一种基于多特征融合卷积神经网络的显著性检测方法。使用多个局部特征增强模块和全局上下文建模模块对卷积神经网络进行增强,利用局部特征增强模块增大特征提取范围的同时,采用全局上下文建模获得特征图的全局信息,有效地抑制了区域内显著而全局不显著的物体对显著性检测的干扰;能够同时提取多尺度局部特征和全局特征进行显著性检测,有效地提升了检测结果的准确性。最后,通过实验对所提方法的有效性进行验证并和其它11种显著性检测方法进行对比,结果表明所提方法能提升显著性检测结果的准确性且优于参与比较的11种方法。With the development of deep learning technology and the prominent performance of con-volutional neural networks in many computer vision tasks,deep saliency detection methods based on convolutional neural networks have become the mainstream methods in saliency detection.However,the convolutional neural network is limited by the size of the convolution kernel,which can only extract features in a small region at the bottom of the network,and cannot detect the objects that are not notable in the region but are globally remarkable.On the other hand,the convolutional neural network can obtain the global information of the image by stacking the convolutional layers,but when the information is transferred from shallow layers to deep layers,it will lead to the loss of information,and stacking too deep will also make the network difficult to optimize.For these reasons,a saliency detection method based on multi-feature fusion convolutional neural network is proposed.In this method,the convolutional neural network is enhanced by several local feature enhancement modules and global context mo-deling modules.Specifically,the local feature enhancement module is used to increase the feature extraction range,and the global information of the feature map is obtained by global context modeling,which effectively suppresses the interference of objects in the region which are notable in the region but not significant in the whole image to the saliency detection.It can also extract multi-scale local features and global features simultaneously for salient detection,which effectively improves the accuracy of detection results.Finally,through experiments,the effectiveness of the proposed method is verified and compared with other 11 saliency detection methods.The results show that the proposed method can improve the accuracy of saliency detection and outperform the other 11 methods involved in the comparison.

关 键 词：显著性检测 多尺度 卷积神经网络 局部特征增强 全局上下文建模 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]