A real-time target detection method based on unmanned aerial vehicle video stream

By improving the network structure of YOLOv5 v5.0 and performing real-time target detection on a local server with powerful computing resources, the problems of insufficient speed and accuracy in real-time target detection of UAVs have been solved, and efficient real-time small target detection has been achieved.

CN115880591BActive Publication Date: 2026-06-09NINGBO UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NINGBO UNIV
Filing Date
2022-11-04
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing real-time target detection methods for UAVs have shortcomings in detection speed and accuracy, especially in detecting small targets in high-resolution UAV aerial images, where the rates of missed detection and false detection are high, and existing models are difficult to meet real-time requirements.

Method used

The network structure of YOLOv5 v5.0 has been improved, including the introduction of spatial-channel attention modules in the backbone network and feature fusion network, the addition of self-attention modules, and the improvement of the CIOU bounding box regression loss function, enabling real-time object detection on a local server with powerful computing resources.

Benefits of technology

It achieves real-time target detection on UAV video streams, with fast detection speed to meet real-time requirements, high detection accuracy, and low false positive and false negative rates. In particular, the detection effect of small targets is significantly improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure QLYQS_1
    Figure QLYQS_1
  • Figure QLYQS_3
    Figure QLYQS_3
Patent Text Reader

Abstract

The application discloses a real-time target detection method based on a UAV video stream, four-point improvement is first carried out on a YOLOv5v5.0 version to obtain an improved YOLOv5 network, then the improved YOLOv5 network is trained to obtain a target detection model, the target detection model is stored in a local server, then a streaming server based on a real-time picture of a UAV is built, the UAV end video stream is pulled through the streaming server, finally the local server obtains the video stream from the streaming server, real-time target detection is carried out on the video stream pulled by the streaming server by using the target detection model, and a target detection result is obtained; the method has the advantages of fast detection speed, the ability to meet real-time detection requirements, high detection precision, and low missing detection rate and false detection rate of small-size targets.
Need to check novelty before this filing date? Find Prior Art