Unlock instant, AI-driven research and patent intelligence for your innovation.

Highway visor inclination identification method based on deep semantic segmentation network and image correction

A technology of semantic segmentation and image correction, applied in the field of image processing, can solve the problems of unavoidable lack or tilt of the visor, hidden danger of road safety protection, etc., and achieve the effect of excellent segmentation quality, simple and efficient method, and robust background interference.

Pending Publication Date: 2022-03-08
西安西光产业发展有限公司
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In practice, due to force majeure reasons such as car accidents and wind damage, the highway visors will inevitably be missing or tilted, which will bring great hidden dangers to road safety protection.

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
  • Highway visor inclination identification method based on deep semantic segmentation network and image correction
  • Highway visor inclination identification method based on deep semantic segmentation network and image correction
  • Highway visor inclination identification method based on deep semantic segmentation network and image correction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0082] (1) Construct the collected image data set

[0083] A high-resolution camera is used to collect image data of shading panels on the expressway. In order to collect rich target content of shading panels, in this example, the angle between the camera and the direction of the road is set to π / 3 when collecting images of shading panels, so that one image can be collected Multiple gobo target images.

[0084] (2) Construct feature extraction backbone network

[0085] A deep residual network is built to extract image features for the collected visor pictures. The overall structure is as follows: image 3 shown.

[0086] The input image is passed through the initial module, where the initial module is three 3×3 convolutions and a maximum pooling layer. The output size after the initial module has been reduced by four times, and the number of output channels is 128.

[0087] After stage 1, stage 1 is composed of three residual units. The residual units are small convolution...

Embodiment 2

[0123] The method for identifying the inclination of an expressway visor based on depth semantic segmentation and image correction in this embodiment consists of the following steps:

[0124] (1) Construct the collected image data set

[0125] A high-resolution camera is used to collect image data of the shading plate on the highway. In order to collect rich target content of the shading plate, in this example, the angle between the camera and the road direction is set to π / 4 when collecting the image of the shading plate.

[0126] (2) Construct feature extraction backbone network

[0127] With embodiment 1.

[0128] (3) Construct a hollow space convolution pooling pyramid module

[0129] With embodiment 1.

[0130] (4) Post-processing module construction

[0131] With embodiment 1.

[0132] (5) Construct the secondary prediction network of points

[0133] 1) Upsampling to extract uncertain points on rough features

[0134] According to the results of the previous rough...

Embodiment 3

[0150] The method for identifying the inclination of an expressway visor based on depth semantic segmentation and image correction in this embodiment consists of the following steps:

[0151] (1) Construct the collected image data set

[0152] A high-resolution camera is used to collect image data of the shading plate on the highway. In order to collect rich target content of the shading plate, in this example, the angle between the camera and the direction of the road is set to π / 5 when collecting the image of the shading plate.

[0153] (2) Construct feature extraction backbone network

[0154] With embodiment 1.

[0155] (3) Construct a hollow space convolution pooling pyramid module

[0156] With embodiment 1.

[0157] (4) Post-processing module construction

[0158] With embodiment 1.

[0159] (5) Construct the secondary prediction network of points

[0160] 1) Upsampling to extract uncertain points on rough features

[0161] According to the results of the previou...

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 relates to a highway visor inclination identification method based on a deep semantic segmentation network and image correction, and the method comprises the steps: inputting a visor image, collected by a high-resolution camera, on a highway into a deep residual neural network model, achieving the pixel-level fine segmentation of a visor region, and carrying out the segmentation of a plurality of visor regions in the same image, the maximum circumscribed polygon is estimated, the optimal circumscribed quadrangle is determined according to the distance between the vertexes, and then shading plate area correction is completed through affine transformation. And finally, estimating the attitude direction of each shading plate by using least squares to realize identification and positioning of the inclined shading plate. Compared with a traditional image segmentation method, the target segmentation model used in the method has better target detail segmentation quality and is very robust to background interference. The method for quickly determining the circumscribed quadrangle through the vertex distance is simple, efficient and suitable for estimating the area shape in a dynamic scene. The effectiveness and superiority of the method are verified by testing measured data.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to image recognition and detection, and in particular to a method for recognizing the inclination of an expressway visor based on depth semantic segmentation and image correction. Background technique [0002] The expressway visor is located in the middle of two opposite roads on the expressway, which can block the direct light from the headlights driving on the diagonally opposite reverse lane, prevent glare and ensure that the driver's eyes are not exposed to strong light; at the same time, it can also help wind flow Orientation to avoid crosswind (wind flow perpendicular to the direction of the expressway) from blowing down and damaging the visor. The installation and deployment of expressway shading panels and the monitoring of defect status play a great role in ensuring the driving safety of expressways. In practice, due to force majeure reasons such as car accidents and wi...

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): G06V20/62G06V10/24G06V10/80G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/25
Inventor 周祚峰吴清泉
Owner 西安西光产业发展有限公司