Real time DSP based identification of surface defects using content-based imaging technique

Abstract

In this paper we propose a technique for the design and development of an automatic visual inspection system for identification of surface defects produced in the steel industry. The proposed DSP based system is implemented using an interconnection of four subsystems: (i) sensing, (ii) data acquisition, (iii) feature (content) extraction and (iv) feature comparison. The system is based on identification of defects using content-based matching of query image with those of database images. The query image is the on-line grabbed image by CCD cameras. The database (off-line) is prepared for all the expected query image by extracting relevant features. An image histogram is used for feature extraction. The computational complexity and storage requirements are reduced by decomposing histogram using wavelet transform. The root mean square (RMS) metric is used for distance calculation.