当前位置>主页 > 期刊在线 > 计算机技术 >

计算机技术2020年24期

基于改进Sobel 算子的语义分割算法
刘清华¹’²,仲臣¹’²,徐锦修¹’²,韩雨辰¹’²
(1. 安徽理工大学 空间信息与测绘工程学院,安徽 淮南 232001; 2. 安徽理工大学 矿山采动灾害空天地协同监测与预警安徽普通高校重点实验室,安徽 淮南 232001)

摘  要:针对常用语义分割算法存在丢失边缘信息导致分割不够精确的问题,通过结合边缘检测算法进行语义分割,有效地改善了分割不准确及边缘模糊的问题。算法采用并行结构,通过边缘检测子网络所提取的边缘特征来对语义分割子网络所提取的语义分割特征进行信息的补充,采用concat 融合两路特征进行卷积操作来获取最终分割结果。实验基于TensorFlow 平台进行,所提出方法相比以往算法在计算速度接近的同时真实值和预测值的交并比上取得了一定提升,增强了分割结果。


关键词:图像分割;边缘检测;深度学习;全卷积神经网络



中图分类号:TP391.41;TP242         文献标识码:A         文章编号:2096-4706(2020)24-0101-05


Semantic Segmentation Algorithm Based on Improved Sobel Operator

LIU Qinghua1,2,ZHONG Chen1,2,XU Jinxiu1,2,HAN Yuchen1,2

(1.School of Space Information and Surveying Engineering,Anhui University of Science and Technology,Huainan 232001,China;2.Key Laboratory of Aviation-Aerospace-Ground Cooperative Monitoring and Early Warning of Coal Mining-Induced Disasters of AnhuiHigher Education Institutes,Anhui University of Science and Technology,Huainan 232001,China)

Abstract:Aiming at the problem of inaccurate segmentation caused by the loss of edge information in common semanticsegmentation algorithms,the problem of inaccurate segmentation and fuzzy edge is effectively improved by semantic segmentationcombined with edge detection algorithm. The algorithm adopts parallel structure. The edge features extracted by the edge detection subnetworkare used to supplement the semantic segmentation features extracted by the semantic segmentation sub-network. The finalsegmentation result is obtained by convolution operation of concat fusion of two features. The experiment is based on TensorFlow platform.Compared with the previous algorithm,the proposed method achieves a certain improvement in the intersection and union ratio of the realvalue and the predicted value,and enhances the segmentation results.

Keywords:image segmentation;edge detection;deep learning;fully convolutional network


基金项目:国家自然科学基金资助项目(41474026)﹔安徽省教育厅资助项目(2018jyxm0192)


参考文献:

[1] LONG J,SHELHAMER E,DARRELL T. Fully ConvolutionalNetworks for Semantic Segmentation [J].IEEE Transactions on PatternAnalysis and Machine Intelligence,2015,39(4):640-651.

[2] CHEN L,PAPANDREOU G,KOKKINOS I,et al.DeepLab:Semantic Image Segmentation with Deep Convolutional Nets,Atrous Convolution,and Fully Connected CRFs [J].IEEE transactions onpattern analysis and machine intelligence,2018,40(4):834-848.

[3] YU F,KOLTUN V. Multi-scale context aggregationby dilatedconvolutions [J/OL].arXiv:1511.07122 [cs.CV].(2015-11-23).https://arxiv.org/abs/1511.07122.

[4] CHEN L,PAPANDREOU G,KOKKINOS I,et al.DeepLab:Semantic Image Segmentation with Deep Convolutional Nets,Atrous Convolution,and Fully Connected CRFs [J].IEEE transactions onpattern analysis and machine intelligence,2018,40(4):834-848.

[5] CHEN L C,PAPANDREOU G,SCHROFF F,et al. Rethinkingatrous convolution for semantic image segmentation [J/OL].arXiv:1706.05587 [cs.CV].(2017-06-17).https://arxiv.org/abs/1706.05587.

[6]BADRINARAYANAN V,KENDALL A,CIPOLLAR. Segnet:A deepconvolutional encoder-decoder architecture forimagesegmentation [J].IEEE transactions on pattern analysisand machineintelligence,2017,39(12):2481-2495.

[7] RONNEBERGER O,FISCHER P,BROX T. U-net:Convolutional networks for biomedical imagesegmentation [C]//InternationalConference on Medical image computing and computerassistedintervention.Cham:Springer,2015:234-241.

[8] CHEN L C,ZHU Y K,PAPANDREOU G,et al. Encoder-Decoder with Atrous Separable Convolution for Semantic ImageSegmentation [C]// Computer Vision-ECCV 2018.Cham:Springer,2018:833-851.https://doi.org/10.1007/978-3-030-01234-2_49.

[9] PASZKE A,CHAURASIA A,KIM S,et al. Enet:Adeepneural network architecture for real-time semanticsegmentation [J/OL].arXiv:1606.02147 [cs.CV].(2016-06-07).https://arxiv.org/abs/1606.02147.

[10] ZHAO H S,SHI J P,QI X G,et al.Pyramid scene parsing network [C]//Proceedings of theIEEE conference on computer visionand patternrecognition.Honolulu:IEEE,2017:6230-6239.

[11] YANG L,WU X Y,ZHAO D W,et al. An improvedPrewitt algorithm for edge detectionbased on noised image [C]// 20114th International Congress on Image and Signal Processing.Shanghai:IEEE,2011:1197-1200.

[12] LI E S,ZHU S L,ZHU B S,et al. An adaptive edgedetectionmethod based on the Canny operator [C]//2009 InternationalConference on Environmental Science and Information ApplicationTechnology.New York:IEEE,2009:465-469.

[13] GAO W,ZHANG X,YANG L,et al. An improved Sobeledge detection [C]//Proceedings of the 3rd IEEE International Conferenceon Computer Science and Information Technology.Chengdu:IEEE,2010:67-71.

[14] 徐辉,祝玉华,甄彤,等. 深度神经网络图像语义分割方法综述 [J]. 计算机科学与探索,2021,15(1):47-59.

[15] 田萱,王亮,丁琪. 基于深度学习的图像语义分割方法综述 [J]. 软件学报,2019,30(2):440-468.

[16] 缪成根,刘琛. 基于改进索贝尔算子的灰度图像边缘检测 [J]. 物联网技术,2020,10(11):37-38+41.

[17] 袁铭阳,黄宏博,周长胜. 全监督学习的图像语义分割方法研究进展 [J]. 计算机工程与应用,2021,57(4):43-54.

[18] 刘丽霞,李宝文,王阳萍,等. 改进Canny 边缘检测的遥感影像分割 [J]. 计算机工程与应用,2019,55(12):54-58+180.


作者简介:刘清华(1997—),男,汉族,山西晋中人,硕士研究生在读,研究方向:图像识别与图像处理。