Tile Surface Defect Detection Based on Improved Deeplabv3+
Aiming at the problems of slow model operation and inaccurate edge feature extraction in tile surface de-fect detection,an improved Deeplabv3+algorithm is proposed.Firstly,the lightweight network MobileNetv2 is used to replace the backbone network of Xception,which can reduce the complexity and the amount of calculation of the model,and improve the speed of algorithm operation.Then,the attention mechanism module(Convolutional Block Attention Module,CBAM)is introduced to make the improved algorithm pay more attention to edge defect informa-tion and small-scale defect information,suppress unnecessary information and improve segmentation accuracy.Ex-perimental results show that the values of MPA,MIou,FPS and Accuracy of the improved network are 96.41％,91.65％,47.44,and 94.88％,respectively,which are 6.33％,7.76％,5.45 and 4.56％higher than those of the model before the improvement,and the segmentation accuracy and detection speed are significantly improved.It fully proves that the improved algorithm provides an effective approach to tile surface defect detection.
万方数据
维普
