A study of hybrid polarization image fusion method for accurate defect detection

Laser Additive Manufacturing (LAM) uses laser as an energy source to enable additive manufacturing of materials which is not limited by the structure of parts and can be used for complex structures, processing, and manufacturing of difficult and thin-walled parts while realizing the manufacture of complex metals. However, some machining defects occurs in the process of LAM would affect the quality of the parts. Thus, accurate detection and feature extraction of the defect location has become an important issue. This paper proposed a hybrid polarization image fusion method for accurate defect detection. This paper is mainly through the convolutional neural network based on the Laplace pyramid and wavelet transform. The experimental results presented that the information entropy of the proposed algorithm is higher than that of the traditional method by 30.5%, the standard deviation is increased by nearly 1 time, and the intersection of the intensity image and polarization feature is realized efficiently. Furthermore, mutual fusion can greatly improve the accuracy and integrity of the target defect detection results, and the similarity reached 80%.