SSD网络算法模型在车辆轴型识别中的应用

Application of SSD Network Algorithm Model in Vehicle Axle Type Recognition

为了在道路超限超载检测中实时准确识别出车辆轴型,提出了一种基于SSD卷积神经网络的车辆轴型检测方法。该模型基于特征图对车辆轴型进行识别,在VGG16-SSD的基础上加入二次训练的策略,得到优化后的SSD算法模型。模型经过优化后,收敛速度加快,损失函数降低了1%,检测能力提升,二轴车辆轴型的识别准确率达89%。优化后的SSD算法模型能够有效识别不同轴型的车辆,该模型能够满足公路非现场检测需要,检测能力和检测精度能满足公路超限超载检测。

Vehicle axle type is an important basis for judging vehicle load.In order to identify vehicle axle type accurately in real-time in road overload detection,a vehicle axle type detection method based on SSD convolution neural network was proposed.In this model,the vehicle axle type was identified based on the feature map,and the optimized SSD algorithm model was obtained by adding secondary training strategy on the basis of vgg16-ssd.After optimization,the convergence speed of the model was accelerated,the loss function(loss)was reduced by 1%,the detection ability was improved,and the recognition accuracy of two axle vehicle axle type reached 89%.The optimized SSD algorithm model can effectively identify vehicles with different axle types.The model can meet the needs of off-site highway detection,and the detection ability and accuracy can be guaranteed,which can be used for highway overload detection.