Coal Flow and Deviation Segmentation Model for Belt Conveyors Based on Improved PIDNet
CUI Fengjian
ZHANG Mei
Abstract：To achieve accurate detection of the coal flow and deviation of belt conveyor,an improved proportional-integral-derivative network(PIDNet)semantic segmentation model was proposed.Firstly,the feature stabilization(FS)module was introduced to dynamically assign higher weights to critical features,reducing information loss caused by continuous downsampling in the network.Secondly,the connection architecture of the parallel aggregation pyramid pooling(PAPP)module was reconstructed,where standard convolutions were replaced with deformable convolution network(DCN),and a lightweight efficient channel attention(ECA)mechanism was embedded.This ensured the integrity of multi-scale features while reducing model parameters count.Finally,the dual-path edge enhancement(DPEE)module was designed,where edge detection and semantic segmentation paths were collaboratively optimized,improving the intersection over union of edge detection.The results demonstrated that the improved PIDNet model achieved a detection speed of 78 frames/s,with the mean intersection over union and mean accuracy of segmentation targets reaching 86.51%and 95.38%,respectively.Compared to the original PIDNet model,the proposed model achieved improvements of 1.79 and 2.21 percentage points.This study provided effective technical support for coal flow intelligent monitoring of belt conveyors.
Keywords：belt conveyorsemantic segmentation modelFS modulePAPP moduleDPEE module
Publication Date：2025-09-20
Online Publishing Date：2025-09-24(First online date of this platform, not the publication date of the document)
Pages：8( 362-369 )