Abstract
An improved model based on YOLOv5s is proposed for the problem that the YOLOv5 network model does not have high localization accuracy when detecting and identifying obstacles at different distances and sizes from the blind, which in turn leads to low accuracy in measuring distances. There are two main core ideas: firstly, a feature scale and a corresponding prediction head are added to YOLOv5 to improve the detection accuracy of small objects on blind paths. Secondly, SK attention mechanism is introduced in the feature fusion part. It can adaptively adjust the perceptual field for feature maps of different scales and more accurately extract objects of different distances and sizes on the blind path, which can improve the accuracy of detection and the accuracy of subsequent distance measurement. It was experimentally demonstrated that the improved YOLOv5 model improved the mAP by 6.29% compared to the original YOLOv5 model based on a small difference in time consumption. And for each category of AP values, the improvement ranged from 2.13% to 8.19%, respectively. The average accuracy of the measured distance from the obstacle at 1.5m to 3.5m from the camera is 98.20%. This shows that the improved YOLOv5 algorithm has good real-time performance and accuracy.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: International Journal of Advanced Computer Science and Applications
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
