Computer Engineering and Applications ›› 2022, Vol. 58 ›› Issue (2): 208-214.DOI: 10.3778/j.issn.1002-8331.2009-0033

• Pattern Recognition and Artificial Intelligence • Previous Articles     Next Articles

Individual Recognition of Dairy Cow Based on Improved SSD Algorithm

XING Yongxin, SUN Youdong, WANG Tianyi   

  1. College of Big Data and Information Engineering, Guizhou University, Guiyang 550025, China
  • Online:2022-01-15 Published:2022-01-18



  1. 贵州大学 大数据与信息工程学院,贵阳 550025

Abstract: To implement the non-contact and high-precision identification of dairy cows in farm environment, an improved SSD(shallow feature module SSD, SFM-SSD) algorithm based on shallow feature module is proposed to solve the problem that SSD(single shot multibox detector) algorithm is not accurate for target recognition. Firstly, the backbone network of the original SSD algorithm is replaced by VGG16 with MobileNetV2. Secondly, the shallow feature module is designed for the shallow feature map of SSD network structure to expand the perception field of shallow feature map and improve the feature extraction ability of shallow feature map for target object. Finally, the [K]-means clustering algorithm is used to reconstruct the region candidate frame to improve the detection accuracy of the algorithm. Experimental results show that the average accuracy of SFM-SSD algorithm is 3.13 percentage points higher than that of traditional SSD algorithm. Meanwhile, the real-time performance of detection is also improved.

Key words: deep learning, target detection, anti residual network, deep separable convolution, feature enhancement module

摘要: 为了实现养殖场环境下无接触、高精度的奶牛个体有效识别,针对SSD(single shot multibox detector)算法识别准确率不高的问题,提出一种基于浅层特征模块的改进SSD(shallow feature module SSD,SFM-SSD)算法。将原始SSD算法的主干网络由VGG16替换为MobileNetV2,以降低网络的运算量,改善检测的实时性;针对SSD网络结构的浅层特征图设计浅层特征模块,扩大浅层特征图的感受视野,提高浅层特征图对目标物体的特征提取能力;利用[K]均值聚类算法重构区域候选框,提高算法的检测精度。实验结果表明:在奶牛个体识别任务中,SFM-SSD算法的平均准确率比原始的SSD算法提升3.13个百分点。同时检测的实时性也得到改善。

关键词: 深度学习, 目标检测, 反残差网络, 深度可分离卷积, 特征增强模块