Segmentation of Main Weld Seam Area Based on MGLNS-Retinex Image Enhancement Algorithm

Abstract

Weld seam defect detection of industrial products is critical for product quality. In the weld seam image, the main weld area only occupies a very small part. For more efficient weld seam defect detection, we propose a segmentation method for the main weld seam area based on MGLNS-Retinex. Specifically, in order to solve the problems of low brightness, uneven illumination, and inconspicuous details of the weld seam image, we first proposed an image enhancement algorithm called Multi Granularity Local Noise Suppression Retinex(MGLNS-Retinex). Then, the threshold value iteration method is adopted to calculate the threshold value used for the binarizing of the weld seam image. Finally, the upper and lower bounds of the main weld seam are obtained by the statistical waveform analysis of the row gray value, based on which the segmentation of the main weld seam area is completed.