Abstract: Automatic fibex laying quality inspection is both time consuming and inadequate,with low defect contrast.In this regard,according to the thermal characteristics of the fibex tow laying process,a platform for fibex tow placement accuracy detection is constructed,and the morphological algorithm is combined with infrared image features of defects for fibex tow placement accuracy detection.Firstly,the top-hat transform of structural elements with the same shape and different scales is carried out,and then the multi-scale bright and dark regions under each group of scales and multi-scale bright and dark details between adjacent groups of scales are extracted to enhance the image;secondly,the multi-directional and multi-scale structural elements and the morphological edge detection algorithm with coefficients self-adjustment are used to obtain the location of filament bundle edges.The experimental results show that the method balances the coordination problem between edge detection accuracy and antinoise performance on top of image noise reduction and image contrast enhancement,and effectively detects the fibex tow placement accuracy with a maximum placement error of no more than 4.53 %.

Key words: image processing, automated fiber placement, morphology, infrared image, defect detecting