Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Traffic sign recognition method based on improved SSD network

A traffic sign recognition and network technology, applied in the direction of character and pattern recognition, biological neural network models, instruments, etc., can solve the problems of insufficient feature information and poor effect, and achieve reduced hardware consumption, high accuracy, and energy saving cost effect

Inactive Publication Date: 2019-09-27
SOUTH CHINA NORMAL UNIVERSITY
View PDF5 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Therefore, due to the front-end network of the traditional SSD network, the feature information extracted from the large-size feature map is not much, resulting in poor detection of small targets.

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 method based on improved SSD network
  • Traffic sign recognition method based on improved SSD network
  • Traffic sign recognition method based on improved SSD network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0031] Such as figure 1 As shown, a traffic sign recognition method based on improved SSD network is provided, including:

[0032] 1: Prepare data

[0033] The data used in this experiment contains 900 road scene images, and there are multiple or one traffic signs on each image. First, set the image to a uniform size, use the upper left corner of the image as the origin, and mark the position of the traffic sign in the image. , size, category and other information.

[0034] (1.1) Convert the data into the Pascal Voc format. There are five folders under the Pascal VOC folder. We use two of them, namely the JPEGImages and Annotations folders. The JPEGImages folder holds the training pictures. And test pictures, making this folder only needs to convert the pictures into JPG format.

[0035] (1.2) The Annotations folder saves the image file in an xml file, which saves the information of the image, including Ground Truth and the type of the image. You can use some automatic anno...

Embodiment 2

[0049] The present embodiment provides a traffic sign detection method based on an improved SSD deep learning network, the method is composed of the following steps:

[0050] 1: Build an optimized SSD network

[0051] (1.1) The traditional SSD network uses the output of the Conv4_3 layer of VGG16 as the first feature map for detection. The improved front-end network uses 34-layer ResNet as the front-end network, and the last feature map of the residual network As the first detection feature map, the dimension of the last feature map of the residual network is controlled at 38*38*512.

[0052] (1.2) Improve the residual network. The convolution kernel of the traditional 34-layer residual network is 3*3. In order to improve the feature extraction effect, two convolution kernels of 3*3 series are added to the penultimate layer. Two 3*3 convolution kernels in series can be equivalent to a 5*5 size convolution kernel in order to extract more comprehensive features

[0053] (1.3) ...

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 discloses a deep learning target detection and recognition algorithm based on an improved SSD. The method comprises the following steps of firstly, modifying a residual error network, using a series convolution kernel method to be equivalent to a large-size convolution kernel, sending the pictures to 34 layers of residual error networks to obtain the feature maps of different sizes, obtaining the feature map of the last layer of the residual error network, and obtaining the feature map of each convolution layer at the same time. The improved SSD network method provided by the invention aims to improve the detection rate of the SSD network on the small targets and realize the detection of the SSD algorithm on the small traffic signs. Due to the fact that the end-to-end network is adopted, the extra storage equipment are not needed, the consumption of the hardware is reduced, the cost is saved, and meanwhile under the improvement of the residual network, the improved SSD network has better accuracy for the small target detection.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, in particular to the field of deep learning traffic sign recognition, in particular to a traffic sign recognition method based on an improved SSD network. Background technique [0002] Deep learning is currently the highest level of machine learning development. As a method of deep learning, convolutional neural network has good results in object recognition, image processing and other fields. For feature extraction, the convolutional neural network has the advantage of automatically learning image features, reducing manual intervention and extracting high-quality features, thus laying a solid foundation for improving the accuracy of image matching. [0003] Image target detection and recognition based on deep learning SSD algorithm network is a major breakthrough in the field of artificial intelligence in the image field, using deep learning methods to detect and recognize images. ...

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): G06K9/00G06N3/04
CPCG06V20/582G06N3/045
Inventor 潘达儒胡武宋晖
Owner SOUTH CHINA NORMAL UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com