基于改进YOLOv8的刨花板表面缺陷检测

打开文本图片集

中图分类号：S777 文献标识码：A 文章编号：2095-2953（2025）06-0036-11

Surface Defect Detection of Particle Board Based on Improved YOLOv8

ZHANG Tao1 YAN Qing-ming² YANG Chun-mei1 （1.CollegeofMechanicalandElectrical Engineering，NortheastForestry University，Harbin Heilongjiang15O40，China; 2.College ofComputerandControlEnginering，NortheastForestry University，Harbin Heilongjiang150o40，China）

Abstract：In view of the problems of insufficient accuracy，a complex structure and por real-time performance of the particleboard surface defect detection algorithm，this paper proposes an improved YOLOv8 lightweight detection algorithm.By introducing the C2F - CT module to enhance the model's understanding and utilization of image information，the detectionaccuracy and efficiency are improved.Meanwhile，SPPF is replaced bythe SPPF-FSC module to optimize feature extraction and reduce redundancy.EMFPN is used to replace PAN-FPN to enhance key feature learninganddetail capture.Finaly，the lightweight detection head LSCBD is introduced tooptimize the localization and classification performance. The experiments show that the improved algorithm achieves 94.4% and 1.8% improvement in mAP index，and the number of model parameters and computation amount are reduced to 1.78M and 5. 6GFLOPs，respectively，which realizes the purpose of saving computational resources and satisfying real-time inspection of the production line while improving detection accuracy.

Keywords：particle board；defect detection；YOLOV8；model lightweight

随着全球经济的持续增长，刨花板作为人造板材的一种，其行业发展速度尤为显著。（剩余16501字）