Research On Road Information Detection Method Based On Deep Learning

Image detection and recognition technology is widely used in numerous spheres of society,thanks to the rapid growth of deep learning.Road information detection,as an important research direction in the disciplines of automatic and assisted driving,has increasing accuracy and speed requirements for image identification and recognition.As a result,an increasing number of academics are opting for the deep learning-based target detection method for detecting road information.This work focuses on the identification and recognition of road information using deep learning technology and traffic photos as the research object.The road information detection data set is self-made,and the network model is established and trained based on Yolo algorithm to realize the detection of traffic signs,vehicles,and pedestrians,in order to address the problems of poor recognition effect and slow detection speed of road information under complex natural environment conditions.The yolov5 network structure has been enhanced,a scale detection head has been introduced to improve the detection effect of small targets,and the FPN feature fusion approach has been improved in this research.The attention mechanism is introduced to the convolution neural network to extract the image features,and the weighted splicing approach is used to keep the low-level network characteristics.This allows the network to pay attention to the crucial information in the image and improve model detection accuracy.Other target detection algorithms are compared to the improved model.The findings suggest that improving network structure improves recall rate,improves small target identification accuracy,and meets the requirements for road information detection.