摘 要：自动泊车系统已经成为高级辅助驾驶系统（ADAS）中的一项重要功能，车辆在泊车过程中时常会出现泊车不到位、与相邻车位中的车辆发生剐蹭等事故。为提升自动泊车的精准性，文章提出了一种实时检测限位器的改进算法SSD-L，通过定位限位器的位置，对车辆的泊车位置进行修正。该方法对原先的 SSD 网络结构进行精简和改进，并使用卡尔曼滤波增加识别的稳定性。在实际泊车场景中的测试结果表明，SSD-L 算法检测限位器的平均精度 (mAP) 较高，为 95%。
关键词：限位器检测；SSD；卡尔曼滤波；ADAS
DOI:10.19850/j.cnki.2096-4706.2022.02.044
基金项目：2020 年山东华宇工学院科技计划项目（2020KJ16）；2021 年山东华宇工学院科技计划项目（2021KJ012）
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2022）02-0174-04
Limiter Detection Method Based on Improved SSD Algorithm
ZHANG Feng
(Shandong Huayu University of Technology, Dezhou 253034, China)
Abstract: Automatic parking system has become an important function of advanced driver assistance system (ADAS). In the process of parking, there are often accidents such as vehicle not parking in place and rubbing with vehicles in adjacent parking spaces. In order to improve the accuracy of automatic parking, an improved algorithm SSD-L for real-time detection of the limiter is proposed, which modifies the parking position of the vehicle by locating the position of the limiter. This method simplifies and improves the original SSD network structure, and uses Kalman filter to increase the stability of recognition. The test results in the actual parking scene show that SSD-L algorithm has a high average accuracy (mAP) of 95% for limiter.
Keywords: limiter detection; SSD; Kalman filter; ADAS
参考文献：
[1] REDMON J，FARHADI A. YOLOv3：An Incremental Improvement [J/OL].arXiv：1804.02767 [cs.CV].[2021-11-08].https://arxiv.org/abs/1804.02767.
[2] LIU W，ANGUELOV D，ERHAN D，et al. SSD：Single Shot MultiBox Detector [C]//Computer Vision–ECCV 2016.Amsterdam：Springer，2016：21-37.
[3] 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.
[4] SIMONYAN K，ZISSERMAN A. Very Deep Convolutional Networks for Large-Scale Image Recognition [J/OL].arXiv：1409.1556 [cs.CV].[2021-12-03].https://arxiv.org/abs/1409.1556.
[5] SANDLER M，HOWARD A，ZHU M L，e t a l . MobileNetV2：Inverted Residuals and Linear Bottlenecks [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City：IEEE，2018：4510-4520.
[6] NEUBECK A，GOOL L V. Efficient Non-Maximum Suppression [C]//18th International Conference on Pattern Recognition （ICPR06).Hong Kong：IEEE，2006：850-855.
[7] KALMAN R E. A New Approach To Linear Filtering and Prediction Problems [EB/OL].1960：35-45.[2021-11-23].https://ieeexplore.ieee.org/document/5311910.
[8] GIRSHICK R. Fast R-CNN [J/OL].arXiv：1504.08083 [cs.CV].[2021-11-23].https://arxiv.org/abs/1504.08083.
[9] ZHENG Z H，WANG P，LIU W，et al. Distance-IoU Loss：Faster and Better Learning for Bounding Box Regression [J/OL].arXiv：1911.08287 [cs.CV].[2021-11-20].https://arxiv.org/abs/1911.08287v1.
[10] EVERINGHAM M，ESLAMI S M A，GOOL L V，et al. The Pascal Visual Object Classes Challenge：A Retrospective [J].International Journal of Computer Vision，2015，111（1)：98-136.
[11] LIN T Y，MAIRE M，BELONGIE S，et al. Microsoft COCO：Common Objects in Context [J/OL].arXiv：1405.0312 [cs.CV].[2021-11-20].https://arxiv.org/abs/1405.0312.
[12] KETKAR N. Stochastic Gradient Descent [M]//KETKAR N.Deep Learning with Python.[S.I.]：Apress，2014：113-132.
作者简介：张凤（1991—），女，汉族，山东临沂人，讲师，硕士研究生，研究方向：图像处理。