Title: Research on Korean translation error text detection method based on machine vision

Authors: Ziyou Zhou

Addresses: School of Asian Languages, Zhejiang Yuexiu University, Shaoxing, 312000, Zhejiang, China

Abstract: Due to the differences and complexity between languages, Korean machine translation still has translation errors, ambiguity and discontinuity, such as translation errors, ambiguity and discontinuity. Therefore, this paper proposes a Korean translation error text detection method based on machine vision. Firstly, the linear array CCD sensor in machine vision is used to collect Korean translation text images. The images are corrected for distortion and processed for greyscale and enhanced through Gabor filters. Then, the images are segmented into multiple candidate frames for recognising the Korean translation text regions. Finally, combining the results of text region recognition, a CNN-Attention model is constructed. The model is then used to input the image to be recognised, extract text features and match with knowledge points to output detection results. The experimental results show that the minimum text recognition rate of this method is 94.8%, the average detection rate is 97.1%, and the minimum detection time is 0.3s.

Keywords: machine vision; Korean translation; error text detection; linear array CCD sensor; Gabor filters; CNN-Attention model.

DOI: 10.1504/IJRIS.2025.148028

International Journal of Reasoning-based Intelligent Systems, 2025 Vol.17 No.4, pp.272 - 280

Received: 18 May 2023
Accepted: 01 Jul 2023
Published online: 15 Aug 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article