Check patentability & draft patents in minutes with Patsnap Eureka AI!

Road surface weather identification method based on road surface semantic segmentation and convolutional neural network

A convolutional neural network, weather recognition technology, applied in neural learning methods, biological neural network models, character and pattern recognition, etc., to achieve the effect of saving network parameters, high recognition accuracy, and reducing redundant operations

Pending Publication Date: 2022-05-24
SOUTH CHINA UNIV OF TECH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] At present, some general image classification problems have more mature solutions, but there are still many challenges in road weather image classification tasks.
For example, there is still room for improvement in the effect of existing image classification models in the field of weather classification; there are many similar features in different weather classifications, such as sunny and cloudy; challenges need to be solved

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
  • Road surface weather identification method based on road surface semantic segmentation and convolutional neural network
  • Road surface weather identification method based on road surface semantic segmentation and convolutional neural network
  • Road surface weather identification method based on road surface semantic segmentation and convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] like figure 1 As shown, this embodiment discloses a road surface weather recognition method based on road semantic segmentation and convolutional neural network, including the following steps:

[0037] S1. Construct a semantic segmentation network, wherein the semantic segmentation network includes an encoder for feature extraction and downsampling and a decoder for fusion encoder multi-scale features and upsampling, and the encoder adopts ResNet18 convolutional neural network network, the decoder adopts the PPM pyramid pooling convolutional neural network, inputs the original image P, and the semantic segmentation network outputs a semantic image S of the same size as the original image P, and the value of each pixel point is equal to the semantic label of the original image P;

[0038] S2, performing road surface binarization processing on the semantic image S, retaining only the part of the semantic image S with the semantic label as the road surface, and outputting ...

Embodiment 2

[0057] This embodiment further discloses the training process of the semantic segmentation network. In order to train the semantic segmentation network, this example adopts the ade20k dataset (see literature: Zhou B, Zhao H, Puig X, et al. Scene parsing through ade20kdataset[C] / / Proceedings of the IEEE conference on computer vision and patternrecognition.2017: 633-641.), the ade20k data set is a scene data set with 20,000 pictures and 150 semantic segmentation categories, of which the road belongs to the sixth category; in order to train the weather recognition convolutional neural network, in this example, Guangdong, The highway surveillance videos in Jiangxi and Liaoning provinces are sampled and video frames are captured, a total of 45,350 pictures are sampled, 5 categories (sunny, cloudy, rainy, foggy and snowy), and all sampled pictures are 7:3 The proportion of the training set and test set.

Embodiment 3

[0059] In this example, network training and testing are performed on a host with 32G memory. The host is configured with an Nvidia1080Ti GPU and an Intel Core i7-6700 CPU, running at a main frequency of 3.41GHz, the operating system is Windows 10, the program development language is python, based on The pytorch deep learning framework, in the collected road weather data set, the image labels are divided into 5 categories, namely sunny, cloudy, rainy, foggy, snowy, using 36380 images for network training, the learning rate is set to 0.001, The network iterates 20 times and uses 8970 images as the test set. like Figure 4 , the weather classification accuracy rate of the method of the present invention is 84.96%, which is 3.62% higher than the accuracy rate of 81.34% of the convolutional neural network ResNet18. again Figure 4 The confusion matrix can be obtained. Except for the classification of cloudy days, the accuracy of the present invention in foggy, rainy and sunny da...

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 pavement weather identification method based on pavement semantic segmentation and a convolutional neural network. The method comprises the following steps: S1, constructing a semantic segmentation network; s2, carrying out binarization processing on the road surface; s3, constructing a weather recognition convolutional neural network; s4, training a semantic segmentation network and a weather recognition convolutional neural network; and S5, collecting expressway pictures in real time, inputting the expressway pictures into the semantic segmentation network and the weather recognition convolutional neural network, and outputting weather types corresponding to the pictures. The invention provides an efficient weather identification method for the highway pavement scene by extracting the pavement area of the picture and learning the features of the whole picture and the pavement area at the same time, and integrating the global features of the whole picture and the local features of the pavement area.

Description

technical field [0001] The invention relates to the technical field of computer vision and pattern recognition, in particular to a road weather recognition method based on road semantic segmentation and convolution neural network. Background technique [0002] Weather recognition is one of the important classification problems in the field of computer vision. The use of digital images to identify weather is helpful in various applications, especially in human-computer interaction. Highway road weather recognition is specific to the weather conditions of the traffic highways to identify whether the weather conditions are bad, and to judge the weather conditions through road images in a timely manner will help to better implement traffic management. [0003] At present, some general image classification problems have mature solutions, but there are still many challenges in road weather image classification tasks. For example, there is still room for improvement in the effect...

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): G06V20/54G06V10/764G06V10/26G06V10/28G06V10/82G06N3/04G06N3/08G06K9/62
CPCG06N3/08G06N3/047G06N3/045G06F18/2431G06F18/2415Y02A90/10
Inventor 张星明傅伦凯林育蓓
Owner SOUTH CHINA UNIV OF TECH
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com