Application of Improved YOLOV3 Algorithm in Part Recognition
-
摘要: 针对工业生产中小型零件存在漏检及识别率不高等问题,结合深度学习基本理论,提出一种改进YOLOV3网络的零件目标识别算法。该算法首先在零件特征融合结构信息中增加了一个特征尺度,进一步融合深层网络与浅层网络的特征信息,以便更好融合零件的位置信息和语义信息。为克服YOLOV3算法中使用K-means聚类对初值不稳定的缺点,根据不同零件类别宽高比,采用K-means++算法对Anchor框重新进行了聚类。最后,在自制常见的六种零件的数据集上,通过实验对该算法进行了验证。结果表明,所提出的改进算法识别效果优于YOLOV3识别效果,在目标识别检测中具有准确率高的优势。Abstract: Aiming at the problems of missed detection and low recognition rate of small and medium-sized parts in industrial production, combined with the basic theory of deep learning, an improved YOLO(You Only Look Once)V3 network part target recognition algorithm is proposed. The algorithm first adds a feature scale to the part feature fusion structure information, and further fuses the feature information of the deep network and the shallow network to better integrate the position information and semantic information of the part. In order to overcome the shortcoming of using K-means clustering in YOLOV3 algorithm that the initial value is unstable, according to the aspect ratio of different parts category, K-means++ algorithm is used to re-cluster the Anchor box. Finally, the algorithm was verified by experiments on data sets of the six self-made common parts. The results show that the recognition effect of the improved algorithm proposed is better than that of YOLOV3, and it has the advantage of high accuracy in target recognition and detection.
-
Key words:
- deep learning /
- YOLOV3 /
- feature fusion /
- part recognition
-
表 1 环境配置表
名称 相关配置 CPU Inter® i5-9300H 内存 8G GPU NVIDIA GTX1650 GPU加速库 CUDA10.2 cuDNN7.6 操作系统 Windows10 数据处理 Python3.7 pytorch1.5 opencv4.2 表 2 相关训练参数
参数名 设定值 Weight-decay 0.000 5 Momentum 0.9 learnrate 0.001 Batch-size 16 Step-size 30 gamma 0.1 -
[1] 郑健红, 鲍官军, 张立彬, 等. 结合深度学习与支持向量机的金属零件识别[J]. 中国图象图形学报, 2019, 24(12): 2233-2242 doi: 10.11834/jig.190127ZHENG J H, BAO G J, ZHANG L B, et al. Metal part recognition based on deep learning and support vector machine[J]. Journal of Image and Graphics, 2019, 24(12): 2233-2242 (in Chinese) doi: 10.11834/jig.190127 [2] 苑津莎, 崔克彬, 李宝树. 基于ASIFT算法的绝缘子视频图像的识别与定位[J]. 电测与仪表, 2015, 52(7): 106-112 doi: 10.3969/j.issn.1001-1390.2015.07.021YUAN J S, CUI K B, LI B S. Identification and location of insulator video images based on ASIFT algorithm[J]. Electrical Measurement & Instrumentation, 2015, 52(7): 106-112 (in Chinese) doi: 10.3969/j.issn.1001-1390.2015.07.021 [3] 王丹, 张祥合. 基于HOG和SVM的人体行为仿生识别方法[J]. 吉林大学学报(工学版), 2013, 43(S1): 489-492WANG D, ZHANG X H. Biomimetic recognition method of human behavior based on HOG and SVM[J]. Journal of Jilin University (Engineering and Technology Edition), 2013, 43(S1): 489-492 (in Chinese) [4] 万源, 李欢欢, 吴克风, 等. LBP和HOG的分层特征融合的人脸识别[J]. 计算机辅助设计与图形学学报, 2015, 27(4): 640-650WAN Y, LI H H, WU K F, et al. Fusion with layered features of LBP and HOG for face recognition[J]. Journal of Computer-Aided Design & Computer Graphics, 2015, 27(4): 640-650 (in Chinese) [5] 陈绪, 陈志澜. 基于迁移学习的零件识别方法研究[J]. 制造业自动化, 2019, 41(8): 81-86+94 doi: 10.3969/j.issn.1009-0134.2019.08.018CHEN X, CHEN Z L. Research on part recognition method based on migration learning[J]. Manufacturing Automation, 2019, 41(8): 81-86+94 (in Chinese) doi: 10.3969/j.issn.1009-0134.2019.08.018 [6] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE, 2014 [7] GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision. Santiago: IEEE Press, 2015 [8] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149 doi: 10.1109/TPAMI.2016.2577031 [9] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016 [10] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[C]//Proceedings of the 14th European Conference on Computer Vision. Amsterdam: Springer, 2016 [11] 余永维, 韩鑫, 杜柳青. 基于Inception-SSD算法的零件识别[J]. 光学 精密工程, 2020, 28(8): 1799-1809YU Y W, HAN X, DU L Q. Target part recognition based Inception-SSD algorithm[J]. Optics and Precision Engineering, 2020, 28(8): 1799-1809 (in Chinese) [12] 陈冠琪, 胡国清, JAHANGIR A S M, 等. 基于改进SSD的多目标零件识别系统研究[J]. 新技术新工艺, 2019(8): 72-76CHEN G Q, HU G Q, JAHANGIR A S M, et al. Research on multi-target parts recognition system based on improved SSD[J]. New Technology & New Process, 2019(8): 72-76 (in Chinese) [13] 司小婷. 基于视觉的零件特征识别与分类方法研究与实现[D]. 北京: 中国科学院大学, 2016SI X T. Research and implementation of parts' feature recognition and classification's method based on machine vision[D]. Beijing: University of Chinese Academy of Sciences, 2016 (in Chinese) [14] 郭斐, 靳伍银, 王猛. 基于改进的Faster R-CNN算法的机械零件图像识别[J]. 机械设计, 2019, 36(9): 113-116GUO F, JIN W Y, WANG M. Image recognition of mechanical parts based on the improved Faster R-CNN algorithm[J]. Journal of Machine Design, 2019, 36(9): 113-116 (in Chinese) [15] 叶佳林, 苏子毅, 马浩炎, 等. 改进YOLOv3的非机动车检测与识别方法[J]. 计算机工程与应用, 2021, 57(1): 194-199YE J L, SU Z Y, MA H Y, et al. Improved YOLOv3 non-motor vehicles detection and recognition method[J]. Computer Engineering and Applications, 2021, 57(1): 194-199 (in Chinese) [16] 景亮, 王瑞, 刘慧, 等. 基于双目相机与改进YOLOv3算法的果园行人检测与定位[J]. 农业机械学报, 2020, 51(9): 34-39+25JING L, WANG R, LIU H, et al. Orchard pedestrian detection and location based on binocular camera and improved YOLOv3 algorithm[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(9): 34-39+25 (in Chinese)