Feature extraction of positron image and imaging algorithm based on 3D convolution operation

Abstract

This study introduces a new positron annihilation based on a 3D imaging test method for foreign matter in a metal-enclosed cavity. When the test method is directly applied to the detection of linear impurities or foreign matter in a dense metal sealed cavity, particularly when the sampling time is short, the image feature cannot be extracted owing to the blurry 3D image. A new 3D-directed convolution algorithm is proposed. First, normalization and 3D filter can eliminate inconsistency of sensors and distractions of metal cavities. Second, the 3D-directed convolution kernel is used to enhance the contrast of edges and foreign matter in the images, thereby extracting foreign matter images. The influence of stray particles with different sizes is eliminated with the corresponding filter, thereby ultimately extracting foreign matter outline. This study can reduce blind trial in image processing and also improve the efficiency and automation of image processing. Results indicate that the proposed method is effective for feature extraction and imaging of foreign matters to process 16-bit original bitmap by 3D convolution operation. Moreover, the resolution of the diameter of the linear foreign matter is above 2 mm. The proposed method provides a new and unique detection technique for the nondestructive testing of foreign matter in a metal-enclosed cavity.