Automated System for Defect Identification and Character Recognition using IR Images of SS-Plates
V. Elanangai1, Vasanth Kishore Babu2
1V. Elanangai, EEE, Sathyabama Institute of Science and Technology, Chennai, India.
2Vasanth Kishore Babu, ECE, Vidya Jyothi Institute of Technology, Hyderabad, India.
Manuscript received on 07 August 2019. | Revised Manuscript received on 15 August 2019. | Manuscript published on 30 September 2019. | PP: 6958-6964 | Volume-8 Issue-3 September 2019 | Retrieval Number: C6009098319/2019©BEIESP | DOI: 10.35940/ijrte.C6009.098319
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Defects on the surface of stainless steel(SS) plates are one of the most important factors affecting the quality of SS plates. Problems of manual defect inspections are lack of accuracy and high time consumption, where early and accurate defect detection is a significant phase of quality control. It is indeed in need to distinguish such abnormalities through computer automated classification systems, which would have a persistent vision of identifying and classifying the above mentioned problem with self-trained classification routine. In this paper, develop a sophisticated routine for defect identification and character recognition on SS plates by considering the multiple features of IR images. The proposed method integrates four steps: (1) defect candidate is detected using a Multi-Scale LoG Weighting; (2) features descriptive of defect shape and texture are extracted; (3) defect objects are classified using a classifier based on SVM-RFE model and (4) the character recognition of SS plate is done using pattern correlation. The output of the anticipated routine is assessed by the metrics: accuracy, sensitivity & specificity. The automated defect identification and classifying routine is compared with ANN, Adaboost and Random Forest (RF) classification methods where the classification result of the anticipated routine outperformed the performance of the previous classification methods.
Keywords: SVM-RFE, GLCM, LNDP, Multi-Scale LoG Weighting, Edge Density Enhancement.
Scope of the Article: Automated Software Design and Synthesis