Feature extraction and image processing for underwater weld with laser vision

Automation and intelligence are the development direction for underwater welding, and real-time sensing and detecting of underwater weld position is a key technology, among which laser vision sensing is a good-prospect detecting method. Noise features of weld image under different water environment and underwater V-groove weld image pre-processing are discussed. And both the application of mean shift algorithms on underwater weld image segmentation and linear Hough transform on extracting image features of underwater weld are studied. Experiment results show, after image enhancement and filtering, laser stripe including weld features could be effectively segmented with mean shift algorithms, and linear Hough transform was suited for precisely extracting V-groove weld feature points on the basis of thinning images.