Traffic sign recognition model training method, traffic sign recognition method and device

A traffic sign recognition and model training technology, applied in the field of artificial intelligence, can solve problems such as increasing the computational complexity of the model, and achieve the effects of fast recognition, reduced loss, and improved accuracy and real-time performance.

Pending Publication Date: 2021-12-10
NANJING UNIV OF POSTS & TELECOMM
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although the accuracy of this method has been improved, this method greatly increases the complexity of model calculations

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Traffic sign recognition model training method, traffic sign recognition method and device
  • Traffic sign recognition model training method, traffic sign recognition method and device
  • Traffic sign recognition model training method, traffic sign recognition method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] Improved real-time traffic sign recognition model training method based on FCOS algorithm, such as figure 1 As shown, the overall model network structure is as follows figure 2 As shown, it specifically includes the following steps:

[0035] Step 1: Initialize and set the positive sample threshold to 0.3, the training epoch to 100, and the batch size to 8, and use ImageNet pre-training weights for initialization, and set the number of pixels on the short side of the input image to 800, and the number of pixels on the long side Less than or equal to 1333.

[0036] Step 2: ResNet-50 is used as the backbone network to extract features from the pictures in the TT100K dataset (Tsinghua Traffic Sign Database), and the CABM module is added to the first and last layers of the ResNet-50 network. After the above operations, The final C3, C4, C5 feature map;

[0037] In the improved ResNet-50 network, the swish activation function is used to replace the ReLu activation functio...

Embodiment 2

[0057] A traffic sign recognition method, which puts the picture to be recognized into the above-mentioned trained model for prediction to obtain the final recognition effect.

[0058] In order to verify the prediction effect of the method proposed in the present invention, four different road scene pictures are randomly selected for recognition, and the recognition effect diagram is as follows image 3 As shown, and the recognition time of each picture is 41ms.

[0059] The model of the present invention adds the CBAM module and adopts the swish activation function in the ResNet-50 network. Experiments show that, under the situation of not affecting the recognition speed, adopting the model of the present invention to carry out traffic sign recognition is comparable to the existing conventional ResNet-50 network. Models are compared. Here, the conventional ResNet-50 network model is used as a comparison example. Specifically, the literature He K, Zhang X, Ren S, et al.Deep Re...

Embodiment 3

[0063] The present invention further provides a traffic sign recognition device, which includes a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the program, the steps of the above method are realized.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention provides a traffic sign recognition model training method, a traffic sign recognition method and a traffic sign recognition device, and the traffic sign recognition method greatly avoids the detection time through the recognition characteristics of a single-stage detection algorithm in a deep learning neural network. According to the traffic sign recognition model training method, on the basis of an FCOS algorithm, the phenomenon of imbalance of positive and negative samples is reduced by introducing an attention mechanism CBAM; a swish function is introduced to avoid a certain degree of feature loss phenomenon; and features of different scales are protected through multi-scale feature fusion. According to the traffic sign recognition method, real-time detection and recognition of traffic signs in a real road scene can be realized, and the problems of small detection object target, natural environment interference, real-time recognition and the like in detection are solved.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and mainly relates to a traffic sign recognition model training method, a traffic sign recognition method and a device. Background technique [0002] Currently, autonomous driving is one of the best ways to solve social problems in the transportation industry. Among them, as an important part of automatic driving, the main purpose of traffic sign recognition is to locate and classify traffic signs encountered during driving, and provide real-time decision support for intelligent transportation systems. Although general object recognition has achieved good results on the PASCAL VOC dataset and the COCO dataset in the past few years, these general object recognition rarely perform well because traffic signs are smaller than objects in general natural scenes. Directly applied to the task of traffic sign recognition, the problems that current traffic sign recognition technology needs...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/241G06F18/253
Inventor 陈哲程艳云
Owner NANJING UNIV OF POSTS & TELECOMM
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap