一种高效的车体表面损伤检测分割算法
作者:
作者单位:

1.福建船政交通职业学院信息与智慧交通学院, 福州 350007;2.福建师范大学数学与信息学院, 福州 350007;3.台湾中山大学资讯工程系, 高雄 80424

作者简介:

通讯作者:

基金项目:

国家自然科学基金(61841701)资助项目;福建省交通运输厅科技基金(201934)资助项目。


Efficient Damage Detection Segmentation Algorithm of Vehicle Body Surface
Author:
Affiliation:

1.College of of Information and Intelligent Transportation, Fujian Chuanzheng Communications College, Fuzhou 350007, China;2.College of Mathematics and Informatics, Fujian Normal University, Fuzhou 350007, China;3.Department of Information Engineering, Sun Yat sen University of Taiwan, Kaohsiung 80424, China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
    摘要:

    车体表面损伤检测是计算机视觉中的经典问题。车体表面损伤检测的主要瓶颈在于图像中损伤实例的不同尺度影响了分割的精度与效率。本文采用单阶段语义分割网络(YOLACT++)进行车体表面的损伤检测,通过结合EfficientNet设计主干网络提高分割效率,并通过改进损失函数优化YOLACT++中目标实例Mask的生成,实验中用深度学习标注实验数据集进行训练测试。实验表明,改进后的YOLACT++降低了Mask生成误差,检测的实时帧率提高到35帧/s,同时也提高了YOLACT++进行实例分割的精度。

    Abstract:

    Car body surface damage detection is a classic problem in computer vision. The main bottleneck of car body surface damage detection lies in the different scales of damage instances in the image, which affects the accuracy and efficiency of segmentation. In this paper, we use a single-stage semantic segmentation network (YOLACT++) for damage detection on the car body surface, combine EfficientNet to design a backbone network to improve segmentation efficiency, and improve the loss function optimization YOLACT++ to generate the target instance Mask in the experiment. Experimental data are marked by deep learning, and results show that the improved YOLACT++ detection frame rate is increased to 35 frame/s, which reduces the mask generation error and improves the instance segmentation accuracy of YOLACT++.

    表 1 不同主干网络YOLACT++模型参数表Table 1 YOLACT++ model parameter table of different backbone networks
    表 5 采用不同基础主干网络的mAP比较表Table 5 Comparison of mAP using different basic backbone networks
    表 2 改进前后YOLACT++梯度损失值对比表Table 2 Comparison table of YOLACT++ gradient loss befor and after improvement
    图1 Sigmoid激励函数曲线Fig.1 Sigmoid activation function curve
    图2 基于改进YOLACT++的车体表面的损伤检测Fig.2 Damage detection of car body surface based on improved YOLACT++
    图3 车体表面损伤目标定位过程Fig.3 Target location process of vehicle body damage
    图4 Mask生成梯度损失率走势对比图Fig.4 Comparison of trend of gradient loss rate generated by the Mask
    图5 不同方法测试查准率对比Fig.5 Comparison of test accuracy of different methods
    图6 不同模型实时性参数对比Fig.6 Comparison of real-time parameters for different models
    表 3 采用不同损失函数的mAP比较表Table 3 Comparison of mAPs using different loss functions
    表 4 不同主干网络的Mask损失值对比表Table 4 Mask loss comparison of different backbone networks
    参考文献
    相似文献
    引证文献
引用本文

林少丹,冯晨,陈志德,朱可欣.一种高效的车体表面损伤检测分割算法[J].数据采集与处理,2021,36(2):260-269

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
历史
  • 收稿日期:2020-07-08
  • 最后修改日期:2020-12-14
  • 录用日期:
  • 在线发布日期: 2021-04-15