Research on License Plate Recognition Method for Complex Scenes in Smart Cities Based on Improved YOLOv7 Algorithm
Aiming at the problems of leakage detection and excessive amount of model parameters of existing license plate detection algorithms in complex environments,a Chinese license plate detection method with improved YOLOv7 is proposed.Firstly,the backbone network of YOLOv7 uses lightweight convolutional GhostConv to reduce the model training parameters.Secondly,CBAM attention mechanism is introduced to improve the feature extraction ability of small target license plate.Then,Head partly replaces part of the convolution using CoordConv,which can perceive the spatial information of the feature map,to improve the detection performance of the model in the scene with large background interference.Finally,the loss function CIoU is replaced with NWD to improve the convergence speed of the algorithm during training.The experimental results show that the improved algorithm improves the detection accuracy of the model while significantly reducing the model parameters and computational volume,and realizes the high-precision detection performance of Chinese license plate detection in complex scenes.
complex scenarioslicense plate detectionYOLOv7attention mechanismloss function
NETL
NSTL
万方数据
