Foreign object debris detection on airport pavement based on YOLOv5-s
To address the challenges posed by the substantial variability of environmental conditions and the di-minished detection accuracy of small targets within optical image-based foreign object debris(FOD)detection on airport pavement,a novel YOLOv5-s-based algorithm for FOD detection is presented.The dark channel pri-or technique is introduced to realize the feature restoration of foggy image data.Additionally,the multi-scale feature extraction,the convolutional block attention module(CBAM),the bidirectional feature pyramid net-work(BiFPN),and a decoupled combination prediction structure are employed to enhance the detection abili-ty of the model.The results indicate that a notable 7.62％enhancement in mean average precision mAP0.5:0.95 following fogged data restoration via defogging.Furthermore,under identical test set conditions,the im-proved algorithm achieves a 7％precision augmentation relative to its predecessor.Notably,the improved al-gorithm demonstrates proficiency in detecting small targets with resolution sizes less than 32 × 32 pixels,exhib-iting a marked improvement in mAP0.5 by 38.40％.The proposed algorithm realizes the high-precision detec-tion of FOD on the airport pavement and provides a new effective method for the construction of FOD real-time detection system.
万方数据
维普
