融合卷积特征的清晰边缘检测研究

doi:10.3778/j.issn.1673-9418.2206085

摘要/Abstract

摘要： 受益于卷积神经网络（CNN），边缘检测性能在多个基准数据集上都已经超过人类水平。但这类算法无法保证边缘的清晰性和定位的准确性。为获取细化清晰、有效抑制背景纹理、定位准确的目标边缘图，提出了一种融合卷积特征的清晰边缘检测算法（FCF）。该算法使用VGG16作为主干网络进行卷积特征提取，将不同阶段的卷积特征上采样后进行特征融合，并通过所设计的细化融合模块（RFB）获得清晰的边缘图。RFB使用多个归一化细化块（GRB）来细化得到的边缘图。此外，为平衡边缘像素和非边缘像素，还提出一个细化骰子损失函数（RD）。在BSDS500数据集上，所提方法将HED、RCF等深度边缘检测器的F-score（ODS）分别提高了2.8%和2.1%；当不使用非极大值抑制（NMS）进行边缘检测评估时，F-score（ODS）、F-score（OIS）分别达到0.801和0.816，超过了其他算法。

关键词: 清晰边缘检测, 融合卷积特征（FCF）, 细化骰子损失（RD）, 卷积神经网络（CNN）

Abstract: Benefiting from convolutional neural networks (CNN), edge detection has surpassed human performance on several benchmark datasets. However, such algorithms cannot guarantee the crispness of edges and the accuracy of positioning. In order to obtain the target edge map that is refined and clear, suppressing background texture effectively, and locating accurately, a crisp edge detection algorithm with fusion of convolutional features (FCF) is proposed in this paper. The algorithm uses VGG16 as the backbone network for convolutional feature extraction, fuses the convolutional features at different stages after upsampling, and obtains a crisp edge map through the refine fusion block (RFB) designed in this paper. RFB uses multiple GroupNorm refine blocks (GRB) to refine the resulting edge map. In addition, to balance edge pixels and non-edge pixels, this paper also proposes a refine dice loss (RD) function. On the BSDS500 dataset, the method proposed in this paper improves the F-score (ODS) of deep edge detectors such as HED and RCF by 2.8% and 2.1%, respectively. When edge detection evaluation is performed without non-maximal suppression (NMS), the F-score (ODS) and F-score (OIS) reach 0.801 and 0.816, respectively, outperforming other algorithms.

Key words: crisp edge detection, fusion of convolutional features (FCF), refine dice loss (RD), convolutional neural networks (CNN)

王兵, 黄刚, 张兴鹏. 融合卷积特征的清晰边缘检测研究[J]. 计算机科学与探索, 2023, 17(9): 2148-2160.

WANG Bing, HUANG Gang, ZHANG Xingpeng. Research on Crisp Edge Detection with Fusion of Convolutional Features[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(9): 2148-2160.

参考文献

[1] WANG S, ZHANG J, HAN T X, et al. Sketch-based image retrieval through hypothesis-driven object boundary selection with HLR descriptor[J]. IEEE Transactions on Multimedia, 2015, 17(7): 1045-1057.
[2] 张萍萍, 李童, 李茹, 等. 一种改进的Sobel图像边缘检测算法及其实现[J]. 电视技术, 2022, 46(5): 42-45.
ZHANG P P, LI T, LI R, et al. An improved Sobel image edge detection algorithm and its implementation[J]. TV Technology, 2022, 46(5): 42-45.
[3] SHEN X Z, ZENG W, GUO Y L, et al. Edge detection algorithm of plant leaf image based on improved Canny[C]//Proceedings of the 6th International Conference on Intelligent Computing and Signal Processing, Xi’an, Apr 9-11, 2021. Piscataway: IEEE, 2021: 342-345.
[4] WANG L, SUN Y. Improved Canny edge detection algori-thm[C]//Proceedings of the 2021 2nd International Confer-ence on Computer Science and Management Technology, Shanghai, Nov 2-14, 2021. Piscataway: IEEE, 2021: 414-417.
[5] SHEN W, WANG X G, WANG Y, et al. DeepContour: a deep convolutional feature learned by positive-sharing loss for contour detection[C]//Proceedings of the 2015 IEEE Confe-rence on Computer Vision and Pattern Recognition, Boston, Jun 7-12, 2015. Washington: IEEE Computer Society, 2015: 3982-3991.
[6] BERTASIUS G, SHI J B, TORRESANI L. DeepEdge: a multi- scale bifurcated deep network for top-down contour detec-tion[C]//Proceedings of the 2015 IEEE Conference on Com-puter Vision and Pattern Recognition, Boston, Jun 7-12, 2015. Washington: IEEE Computer Society, 2015: 4380-4389.
[7] ZHANG X, CHEN S, ZHU P, et al. Spatial pooling graph convolutional network for hyperspectral image classifica-tion[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-15.
[8] XU Z, YU H, ZHENG K, et al. A novel classification framework for hyperspectral image classification based on multiscale spectral-spatial convolutional network[C]//Proce-edings of the 2021 11th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing, Amst-erdam, Mar 24-26, 2021. Piscataway: IEEE, 2021: 1-5.
[9] DIVYA S, ADEPU B, KAMAKSHI P. Image enhancement and classification of CIFAR-10 using convolutional neural networks[C]//Proceedings of the 2022 4th International Conference on Smart Systems and Inventive Technology, Tirunelveli, Jan 20-22, 2022. Piscataway: IEEE, 2022: 1-7.
[10] TAO Y, ZHAO Z Y, ZHANG J, et al. Low-altitude small-sized object detection using lightweight feature-enhanced convolutional neural network[J]. Journal of Systems Engin-eering and Electronics, 2021, 32(4): 841-853.
[11] 王玲敏, 段军, 辛立伟. 引入注意力机制的YOLOv5安全帽佩戴检测方法[J]. 计算机工程与应用, 2022, 58(9): 303-312.
WANG L M, DUAN J, XIN L W. YOLOv5 helmet wear detection method with introduction of attention mechanism[J]. Computer Engineering and Applications, 2022, 58(9): 303-312.
[12] 杨永波, 李栋. 改进YOLOv5的轻量级安全帽佩戴检测算法[J]. 计算机工程与应用, 2022, 58(9): 201-207.
YANG Y B, LI D. Lightweight helmet wearing detection algorithm of improved YOLOv5[J]. Computer Engineering and Applications, 2022, 58(9): 201-207.
[13] 栗俊杰, 毛鹏军, 淡文慧, 等. 基于YOLOv2-Tiny的无人机火灾检测与云台跟踪研究[J]. 消防科学与技术, 2022, 41(1): 5-11.
LI J J, MAO P J, DAN W H, et al. Research on UAV fire detection and PTZ tracking based on YOLOv2-Tiny[J]. Fire Science and Technology, 2022, 41(1): 5-11.
[14] YIN R, CHENG Y, WU H, et al. FusionLane: multi-sensor fusion for lane marking semantic segmentation using deep neural networks[J]. IEEE Transactions on Intelligent Trans-portation Systems, 2022, 23(2): 1543-1553.
[15] WANG R, QIU K. Fine-grained remote sensing semantic segmentation method under multi-stage supervision learn-ing[C]//Proceedings of the 2022 14th International Confer-ence on Measuring Technology and Mechatronics Automation, Changsha, Jan 15-16, 2022. Piscataway: IEEE, 2022: 361-365.
[16] XIE S, TU Z. Holistically-nested edge detection[C]//Proce-edings of the 2015 IEEE International Conference on Computer Vision, Santiago, Dec 7-13, 2015. Washington: IEEE Computer Society, 2015: 1395-1403.
[17] LIU Y, CHENG M M, HU X, et al. Richer convolutional features for edge detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(8): 1939-1946.
[18] WANG Y, ZHAO X, HUANG K. Deep crisp boundaries[C]//Proceedings of the 2017 IEEE Conference on Com-puter Vision and Pattern Recognition, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 1724-1732.
[19] ARBELáEZ P, MAIRE M, FOWLKES C, et al. Contour detection and hierarchical image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(5): 898-916.
[20] SILBERMAN N, HOIEM D, KOHLI P, et al. Indoor segmentation and support inference from RGBD images[C]//LNCS 7576: Proceedings of the 12th European Confe-rence on Computer Vision, Florence, Oct 7-13, 2012. Berlin, Heidelberg: Springer, 2012: 746-760.
[21] 黄胜, 冉浩杉. 基于语义信息的精细化边缘检测方法[J]. 计算机工程, 2022, 48(3): 204-210.
HUANG S, RAN H S. Refined edge detection method based on semantic information[J]. Computer Engineering, 2022, 48(3): 204-210.
[22] 岳欣华, 邓彩霞, 张兆茹. BP神经网络与形态学融合的边缘检测算法[J]. 哈尔滨理工大学学报, 2021, 26(5): 83-90.
YUE X H, DENG C X, ZHANG Z R. BP neural network fuse with morphology edge detection method[J]. Journal of Harbin University of Science and Technology, 2021, 26(5): 83-90.
[23] 申嘉锡, 齐华, 王晨. Canny算子对图像边缘检测的一种改进[J]. 现代计算机, 2022, 28(3): 46-49.
SHEN J X, QI H, WANG C. The improvement of tradi-tional Canny image edge detection algorithm[J]. Modern Computer, 2022, 28(3): 46-49.
[24] HE J, ZHANG S, YANG M, et al. BDCN: bi-directional cascade network for perceptual edge detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(1): 100-113.
[25] DENG R, SHEN C, LIU S, et al. Learning to predict crisp boundaries[C]//LNCS 11210: Proceedings of the 15th Euro-pean Conference on Computer Vision, Munich, Sep 8-14, 2018. Cham: Springer, 2018: 570-586.
[26] CAO Y J, LIN C, LI Y J. Learning crisp boundaries using deep refinement network and adaptive weighting loss[J]. IEEE Transactions on Multimedia, 2021, 23: 761-771.
[27] HUAN L, XUE N, ZHENG X, et al. Unmixing convol-utional features for crisp edge detection[J]. IEEE Transa-ctions on Pattern Analysis and Machine Intelligence, 2022,44(10): 6602-6609.
[28] MARTIN D R, FOWLKES C C, MALIK J. Learning to detect natural image boundaries using local brightness, color, and texture cues[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(5): 530-549.
[29] DOLLAR P, ZITNICK C L. Fast edge detection using structured forests[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(8): 1558-1570.
[30] LONG J, SHELHAMER E, DARRELL T. Fully convolu-tional networks for semantic segmentation[C]//Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition, Boston, Jun 7-12, 2015. Washington: IEEE Computer Society, 2015: 3431-3440.
[31] PASZKE A. Automatic differentiation in PyTorch[C]//Proceedings of the 2017 31st Conference and Workshop on Neural Information Processing Systems, Long Beach, Dec 4-9, 2017.
[32] HALLMAN S, FOWLKES C C. Oriented edge forests for boundary detection[C]//Proceedings of the 2015 IEEE Con-ference on Computer Vision and Pattern Recognition, Boston, Jun 7-12, 2015. Washington: IEEE Computer Soc-iety, 2015: 1732-1740.
[33] BERTASIUS G, SHI J B, TORRESANI L. High-for-low and low-for high: efficient boundary detection from deep object features and its applications to high-level vision[C]//Proceedings of the 2015 IEEE International Conference on Computer Vision, Santiago, Dec 7-13, 2015. Washington: IEEE Computer Society, 2015: 504-512.
[34] BAO S S, HUANG Y, XU G Y, et al. Bidirectional multiscale refinement network for crisp edge detection[J].IEEE Access, 2022, 10: 26282-26293.
[35] 杨红菊, 王昱蓉. FMLED: 细粒度级多尺度特征表示的轻量级边缘检测方法[J]. 小型微型计算机系统, 2023, 44(4): 812-817.
YANG H J, WANG Y R. Fine-grained multi-scale feature representation lightweight network for edge detection[J]. Journal of Chinese Computer Systems, 2023, 44(4): 812-817.
[36] 郑恩壮, 钟宝江. 各向异性的多尺度边缘检测算法[J]. 激光与光电子学进展, 2022, 59(4): 282-290.
ZHENG E Z, ZHONG B J. Anisotropic multi-scale edge detection algorithm[J]. Laser & Optoelectronics Progress, 2022, 59(4): 282-290.