Manhole Cover Disease Detection Based on Multi-Scale Feature Fusion and Attention Mechanism
HUANG Hui
LÜ Yonghe
CHEN Yuejun
Abstract：A method for identifying manhole cover diseases based on multi-scale feature fusion and attention mechanism is proposed to address the issues of damage,settlement,and displacement of urban lighting manhole covers.This method constructs a multi-level feature perception and fusion system through three key technological innovations:Design a hybrid sliding Transformer module in the feature extraction stage to enhance feature perception in complex backgrounds;Build a generalized feature pyramid network to enhance the representation ability of multi-scale features through cross layer and cross scale connections;Design a dynamic detection head to achieve multi-dimensional attention regulation.At the same time,a well cover image dataset with wide coverage was constructed,and the effective ness and generalization ability of the proposed method were verified by comparing it with other methods in multiple scenarios.
Keywords：urban lighting manhole coverdisease detectionmulti scale feature fusionattention mechanism
Publication Date：2026-01-25
Online Publishing Date：2026-03-05(First online date of this platform, not the publication date of the document)
Pages：8( 72-79 )