基于全景图像的车位检测方法研究

Parking Space Detection Method Based on Panoramic Images

针对自动泊车过程中车位检测网络复杂、边缘部署困难等问题,提出了一种基于全景图像的车位检测方法。使用Ghost模块对参数进行量化处理;引入EIoU损失函数对检测尺度进行裁剪,以降低计算成本;添加SimSPPF模块替换SPPF模块,以提高计算效率和目标检测能力;对检测到的车位角点进行配对并推理出完整停车位。实验结果表明,该模型在保证检测精度的前提下大幅降低了网络复杂度,在检测性能和部署难度上优于以往常见方法。

A parking space detection method based on panoramic images is proposed to address the challenges of complex networks and difficulties in edge deployment during the automated parking process.Ghost module is adopted to quantize the parameters and the EIoU loss function is introduces to crop the detection scale,reducing computational costs.Additionally,the SimSPPF module is added to replace the SPPF module to enhance computational efficiency and accurate object detection capabilities.The detected parking corners are matched and the complete parking space can be deduced.The experimental results show that the proposed model greatly reduces the network complexity while ensuring the detection accuracy,and is superior to the common methods in detection performance and deployment difficulty.