YOLO ground mark detection method and device based on perspective downsampling and storage medium

A technology of ground markings and detection methods, which is applied in the direction of neural learning methods, instruments, biological neural network models, etc., can solve problems such as traffic accidents, low real-time performance, and reduced recognition accuracy, so as to improve the network structure and eliminate perspective Effects that deform and reduce image size

Pending Publication Date: 2021-09-03
SHANDONG UNIV
View PDF10 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Drivers may fail to notice ground signs due to tree shadow occlusion, light changes, etc., or may not understand the specific meaning of a landmark, etc., leading to serious traffic accidents and affecting normal traffic order
[0004] Most of the existing ground guidance sign recognition methods are based on the traditional binarization method and SVM in machine learning. The real-time performance is not high, and the accuracy of recognition is not high when encountering signs such as blurring, lighting changes, and shadow occlusion. significantly reduce

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
  • YOLO ground mark detection method and device based on perspective downsampling and storage medium
  • YOLO ground mark detection method and device based on perspective downsampling and storage medium
  • YOLO ground mark detection method and device based on perspective downsampling and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0067] A YOLO ground sign detection method based on perspective downsampling. Ground signs refer to various signs located on the road plane, mainly guiding arrow signs, and mainly study five types of common signs, including going straight or turning right, going straight or left To turn, go straight, turn left, and turn right, in order to display the categories concisely and intuitively, they are represented by SorR, SorL, S, L, and R respectively, such as figure 1 shown, including the following steps:

[0068] (1) Video collection, frame screening, annotation production and construction of data sets

[0069] The road images are acquired and labeled in real time by the on-board camera installed in front of the vehicle, and a data set is constructed. The road images in the data set are divided into training set, test set and verification set.

[0070] (2) Perspective Downsampling

[0071] The features of the ground turning signs are simple, and the pixels in the area where th...

Embodiment 2

[0088] According to a kind of YOLO ground mark detection method based on perspective down-sampling described in embodiment 1, its difference is:

[0089] Train the YOLO target detection model, the specific implementation steps include:

[0090]A. The YOLO network borrows from the GoogLeNet classification network structure. As long as the input image is detected once, the position of all objects in the image and the probability of their category will be obtained. Input the road image in the training set into the YOLO target detection model, and divide the road image into S×S grids. If an object, that is, the center point of the real frame of a certain sign falls into a certain grid, then the grid Responsible for predicting the object, each grid predicts B bounding boxes and the confidence score of the bounding box, the confidence is the probability that each bounding box contains the object, specifically including: First, the possibility that the bounding box contains the targe...

Embodiment 3

[0106] A computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the YOLO ground landmark detection method based on perspective downsampling in embodiment 1 or 2 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 relates to a YOLO ground mark detection method and device based on perspective downsampling and a storage medium, and the method comprises the steps of obtaining an image in front of a vehicle through a vehicle-mounted camera, selecting an image region of interest for perspective downsampling, and reducing the near resolution of a road image; secondly, improving a YOLOv3-tiny target detection network, adding a convolutional layer to strengthen shallow layer features, and improving the small target characterization capability; changing the feature pyramid fusion scale, and adjusting prediction output to 26 * 26 and 52 * 52 suitable for the landmark size; and finally, testing a self-built multi-scene data set, so that the accuracy is improved to 99% from 78%, and the size of the model is reduced to 8.3 MB from 33 MB. When a vehicle runs on a road with a ground sign, the ground steering sign can be accurately detected in real time, the robustness is high, the small target detection precision is higher, and the invention is easy to deploy on low-end embedded equipment.

Description

technical field [0001] The invention relates to a YOLO ground mark detection method, equipment and storage medium based on perspective down-sampling, and belongs to the technical field of computer vision image processing. Background technique [0002] At present, companies such as Google, Baidu, and Tesla are conducting research on unmanned vehicles, and some unmanned test vehicles have entered the stage of actual road testing. The domestic autonomous driving industry continues to rise, and unmanned taxis, unmanned buses, and unmanned delivery have reached a new level. It is believed that in the near future, driverless cars will enter the lives of the public and change the way people travel. [0003] In the field of intelligent transportation, the research on ground traffic signs is mainly lane line recognition, and the research on ground guide sign recognition is less. Landmark detection can accurately locate and recognize the guiding signs on the lane in real time, and s...

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 & AuthorityApplications(China)
IPC IPC(8): G06K9/00G06K9/20G06K9/32G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/047G06F18/23213G06F18/2431G06F18/2415G06F18/214
Inventor陈辉李玉珍
OwnerSHANDONG UNIV