Truck license plate recognition method based on extreme learning deep network fusion model

Abstract

The invention discloses a truck license plate recognition method based on an extreme learning deep network fusion model, comprising the following steps: using a highway monitoring camera to obtain a truck image, using a deformable component model for the truck image, and positioning the truck license plate, Based on the proportional segmentation method, character segmentation is performed on the truck license plate obtained by positioning, and a character image set of the truck license plate is constructed; feature extraction is performed on the character image of the vehicle license plate; feature extraction is performed on the character image of the vehicle license plate; Extract features from images; build a deep network fusion model based on extreme learning; use the trained deep network fusion model based on extreme learning to identify and classify truck license plates on the input vehicle license plate image. The advantage is that the performance of the present invention is superior to the traditional HOG+SVM, extreme learning InceptionV3 model, extreme learning XceptionV3 model and extreme learning NASNet model, and its recognition rate reaches 98.18%.

Description

technical field [0001] The patent of the invention relates to the field of intelligent transportation and intelligent high-speed research, and can be applied to various traffic scenarios, such as: traffic law enforcement system, parking management system, vehicle detection system, traffic guidance system, highway inspection system, vehicle dispatching system and highway scenes Various application scenarios such as smart toll collection system for trucks. Background technique [0002] License plate recognition technology is an application of computer video image recognition technology in vehicle license plate recognition. It is a technology that can detect vehicles on the monitored road and automatically extract vehicle license plate information (including Chinese characters, English letters, Arabic numerals and colors) for processing. License plate recognition is one of the important components of modern intelligent transportation systems, and it is widely used. Based on d...

AI Technical Summary

Problems solved by technology

However, in the convolutional neural network, the traditional feed-forward neural network uses an iterative algorithm of gradient descent to adjust the weight parameters, which makes the learning speed of the convolutional neural network slow, the calculation time increases, the learning rate is difficult to determine, and it is easy to fall into a local minimum. , but also prone to over-training, causing generalization performance to decline

These defects become the bottleneck restricting the wide application of iterative algorithm feed-forward neural network

Images