基于YOLOv5s轻量化改进的LCD缺陷检测方法OA

In view of the current situation of the low efficiency and accuracy in the LCD detection,a lightweight improvement model of YOLOv5s has been proposed in the present paper,which is used to detect and identify the defects existing in LCD.The Nearest is replaced by improving the upsampling CARAFE operator,which is compared by modifying the two parameters of kencoder and kreassembly;at the same time,the CBAM focus mechanism is added,which can pay more attention to the target area feature information model to enhance the recall rate;finally,the C3 is replaced by C3_Ghost with lightweight design to achieve the reduction of the model size,the number of parameters and transport.The experimental results show that on the basis of the original model,the precision of YOLOv5s algorithm is improved by 2.1%,the recall rate is improved by 5.4%,the average precision of the model reaches 88.8%,which is 2.1%higher than that without improvement,and the parameter amount and calculation amount are reduced by 15.6%and 20.9%respectively.And the model size is reduced by 14.6%.Over all,the improved algorithm model is more lightweight,the model MB is reduced and the number of parameters and calculations are relatively reduced,so it is convenient for the deployment of low computing power hardware,and also provides a certain technical reference for LCD factory intelligent detection technology.