| Autor: |
N. C. Dayananda Kumar, K. V. Suresh, R. Dinesh |
| Rok vydání: |
2018 |
| Předmět: |
|
| Zdroj: |
Data Analytics and Learning ISBN: 9789811325137 |
| DOI: |
10.1007/978-981-13-2514-4_16 |
| Popis: |
Innovation anddevelopment of computer vision algorithms finds real-time application inautomated evaluation of optical mark sheets. Most of the existing methods are template specific and requires parameter tuning based on the given template. Developing robust and low-cost solutions for optical mark recognition (OMR) is still a challenging problem. The need for parameter less and computationally efficient method that works in real time adds on to the existing challenges. To address these problems, robust and computationally efficient parameter-less OMR method is proposed. The main contributions of this paper are dynamic localization of optical mark region that makes the algorithm generic and independent of templates. Accuracy is improved by implementing some of the heuristics approaches which identifies marking regions that are missed during preprocessing. Marked regions are accurately detected by computing local threshold instead of global classification margin. The algorithm is evaluated on various OMR templates obtained from different scanners and smartphones. Experimental results depict the efficiency of the proposed method with the error rate less than 2% by processing 150 sheets per minute. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
