Research on LCD Defect Detection Algorithm Based on YOLOv8s

Abstract

Aiming at the problems of small LCD display defects and low contrast, which are easy to be confused with background and lead to unsatisfactory detection effect, a liquid crystal display defect detection algorithm based on YOLOv8s was proposed. Firstly, a new CBLGhost module is pro-posed to facilitate the design core of GhostNet network, which uses the operation of convolution and linear change to generate the feature map, effectively reducing the computing resources re-quired by the model. Secondly, the HorNet module is introduced into the neck network to realize the modeling of high-order spatial interaction, and improve the recognition ability of the model for tiny features. Finally, CFNet module is introduced to balance the proportion of parameters between backbone network and fusion module network, so as to reduce the number of algorithm parameters and improve the detection speed of the algorithm. The experimental results on the self-made LCD defect data set show that the proposed algorithm can improve the detection ac-curacy without sacrificing FLOPs. Compared with the original algorithm, the accuracy is signif-icantly improved, with mAP reaching 93.7%, an increase of 3.8%. Compared with the mainstream target detection algorithms, the results show that the proposed algorithm has better performance in detecting the display defects of LCD.