Method for identifying traffic lights based on YOLOv5 model

The invention discloses a traffic signal lamp identification method based on a YOLOv5 model, and the method employs a K-means clustering algorithm to determine the parameters of a target frame of a target object, adds an attention mechanism in a head target detection head, and changes an up-sampling...

Full description

Saved in:
Bibliographic Details
Main Authors LIU WEI, ZHAO YANBO, HUA SHANG, GUO HAOYU, WANG YUYANG, ZHAO HUANYU
Format Patent
LanguageChinese
English
Published 01.08.2023
Subjects
CALCULATING
COMPUTING
COUNTING
PHYSICS
Online AccessGet full text

Cover

More Information
Summary:The invention discloses a traffic signal lamp identification method based on a YOLOv5 model, and the method employs a K-means clustering algorithm to determine the parameters of a target frame of a target object, adds an attention mechanism in a head target detection head, and changes an up-sampling mode to improve the detection precision. Comprising the steps of shooting image data of a traffic signal lamp at an intersection on a public road, performing fuzzy processing on part of pictures to obtain a data set under the condition of unclear view in rainy and foggy days and the like, performing data labeling on the pictures of the data set in an xml format, converting a file into a txt file which can be recognized by YOLO, and storing the txt file in a database; 80% of the self-made data set is selected as a training set. The improved YOLOv5 model is trained, and a test set is put into the improved network for detection, so that a good effect can be obviously achieved. With gradual research of automatic drivi
Bibliography:Application Number: CN202310408148