A small-scale target detection method, system and storage medium

By integrating a pixel rearrangement feature extraction module and a feature rearrangement network into the YOLOv8 model, the problem of low accuracy in small-scale target detection is solved, and the accuracy and efficiency of detection are improved.

CN118365865BActive Publication Date: 2026-01-27QUANZHOU INST OF EQUIP MFG +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410767180.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-14
Publication Date
2026-01-27
Estimated Expiration
2044-06-14

AI Technical Summary

Technical Problem

Existing target detection models have weak feature representation capabilities when detecting small-scale targets, resulting in low detection accuracy. Furthermore, they are prone to losing the location information of small-scale targets during network deepening, increasing the false negative and false positive rates.

Method used

The YOLOv8 model is adopted, integrating the pixel rearrangement feature extraction module PRFE and the feature rearrangement network FR-Neck. By using feature fusion and contextual information to assist in the understanding of local features, the detection capability of small-scale targets is improved.

Benefits of technology

It effectively reduced the false negative and false positive rates for small-scale targets, and enhanced the model's ability to locate and classify small-scale targets.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118365865B_ABST
    Figure CN118365865B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of computer vision, in particular to a small-scale target detection method and system and a storage medium, S1: acquiring image data sets; S2: integrating a pixel rearrangement feature extraction module PRFE in a backbone network of a YOLOv8 model, replacing a neck network with a feature rearrangement network FR-Neck, setting three feature maps of different scales as detection layers of an input detection head, and obtaining a small-scale target detection model based on pixel rearrangement and feature rearrangement; S3: inputting image data in a training set into the small-scale target detection model based on pixel rearrangement and feature rearrangement, continuously iteratively training, and obtaining a weight file of small-scale target detection; S4: loading the weight file into the small-scale target detection model based on pixel rearrangement and feature rearrangement to obtain a target detection model; and S5: inputting an image to be detected into the target detection model and outputting corresponding target detection results; the model's detection capability for small-scale targets is effectively improved.
Need to check novelty before this filing date? Find Prior Art