Low-resolution small target detection method and system based on improved YOLO model
By improving the YOLO model and utilizing CSHM, AIoU, and IEDH techniques, the problems of low accuracy and frame rate in low-resolution small target detection were solved, achieving efficient detection in complex scenes.
CN118736205BActive Publication Date: 2026-07-24NANJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2024-07-30
- Publication Date
- 2026-07-24
Smart Images
Figure CN118736205B_ABST
Abstract
The application discloses a low-resolution small target detection method and system based on an improved YOLO model, and the method comprises the following steps: collecting low-resolution small target images and constructing a data set; constructing a channel space mixed feature highlighting module CSHM by using the low-resolution characteristics of the data set images, applying the channel space mixed feature enhancement module to the connection between the backbone network and the neck network of YOLOv7tiny; further proposing an adaptive intersection over union calculation method AIoU according to the small target characteristics in the images, constructing an implicit efficient decoupling head structure IEDH, using the AIOU as the intersection over union calculation method in YOLOv7tiny, and decoupling the output feature map of the IEDH; finally, inputting the low-resolution small target images into the improved network to obtain low-resolution small target detection results; and the application can detect low-resolution small targets in real time with high precision.
Need to check novelty before this filing date? Find Prior Art