结合双金字塔特征融合与级联定位的车牌检测

License Plate Detection Using Siamese Feature Pyramid and Cascaded Positioning

为了解决复杂环境中不同因素干扰车牌检测精确度的问题,提出了一种基于双金字塔特征融合的复杂环境下车牌检测算法。通过采用Mish激活函数的残差网络(ResNet101-M)对输入图像进行初级特征提取;在传统特征金字塔网络(feature pyramid network,FPN)的基础上,提出了一种改进的双金字塔特征融合网络(siamese feature pyramid network,SFPN)。被提取的初级特征被送入该网络进行多层特征融合。融合后的特征被送入基于形状先验的锚点设置网络来确定感兴趣区域。将所生成的感兴趣区域送入级联定位网络从而得到准确的车牌检测结果。实验结果表明,该算法在AOLP与CCPD车牌数据集上均能够有效提升检测性能。

In order to solve the problem of license plate detection in complex environment,a novel license plate detection algorithm using siamese feature pyramid and cascaded positioning is proposed.Firstly,the original features of the images are extracted by the residual network with the Mish activation function.At the same time,a siamese feature pyramid network(SFPN)is proposed in this paper.The extracted features are sent to the siamese feature pyramid network for multilevel feature fusion.Then,a region proposal network based on the prior of shape is used to generate the regions of interest.Finally,the regions of interest are fed to the cascaded positioning network to obtain the license plate location.The experimental results show that the proposed algorithm effectively improves the performance of license plate detection on both AOLP and CCPD datasets.