A method for real-time detection and recognition of traffic lights

Abstract

The invention discloses a method for real-time detection and identification of traffic signal lamps, which belongs to the field of detection and identification of traffic signal lamps. The invention is proposed aiming at the problem that the YOLOv4 algorithm is not sensitive to the detection of small targets, which results in low detection accuracy of traffic signal lights. A shallow feature enhancement mechanism is proposed, which fuses the two shallow features in different stages of the feature extraction network with the high-level semantic features obtained after two upsamplings, increases the scale of the two detection layers, and improves the network's ability to detect small objects. Positioning and color resolution capabilities; introduce the uncertainty prediction mechanism of the bounding box, model the output coordinates of the predicted bounding box, add the Gaussian model to calculate the uncertainty of the coordinate information, and improve the reliability of the predicted bounding box. Using the LISA traffic signal data set to conduct detection and recognition experiments, the AUC value of the improved YOLOv4 algorithm in the detection experiment is 97.58%, which is 7.09% higher than that of VIVA; the average accuracy of the improved YOLOv4 algorithm in the recognition experiment is 82.15%, which is higher than the original YOLOv4 algorithm. 2.86%. The improved YOLOv4 algorithm improves the detection and recognition accuracy of traffic lights.

Description

technical field [0001] The invention relates to a real-time detection and identification method for traffic signal lights, and belongs to the field of traffic signal light detection and identification. Background technique [0002] Traffic light detection is one of the important branches in the field of intelligent driving, and its recognition accuracy directly affects the safety of intelligent driving. The traffic light detection system often uses industrial cameras to collect road condition information, and the background of traffic light images collected in actual traffic scenes is complex The signal lights in the image occupy fewer pixels, and the feature structure is sparse, which increases the difficulty of the algorithm to identify. Therefore, it is of great significance to study more effective small target detection algorithms for the problem of traffic light detection. [1-2] . [0003] Traffic signal detection has experienced several years of development, and ther...

AI Technical Summary

Problems solved by technology

[0008] In order to solve the problem that the YOLOv4 algorithm used in the existing traffic signal light detection is not sensitive to small target detection, resulting in low detection accuracy of traffic signal lights, and further provides a method for real-time detection and identification of traffic signal lights

Images