A Hybrid Approach to Detect and Identify Text in Picture

Abstract

In order to create computer systems that can automatically read text from images or pictures, researchers focus on detecting and recognizing text in images. This issue is particularly difficult because images often have complicated backgrounds and a wide range of properties, including color, size, shape, orientation, and texture. Our proposed approach is based on morphology, which consists of a dilation and erosion process to extract text and recognize black-and-white text areas that contain document text or images. This suggested approach has been investigated for its ability to automatically identify text aligned with text pictures, such as store names, street names, banners, and posters. The design, application, and outcomes of the device's experiments are covered in this manuscript using Optical Character Recognition (OCR) Tesseract standards and the optimized OCR Tesseract. Our result shows that the optimized OCR Tesseract performs much better compared to the standard. Image preprocessing and text processing modules comprise this device's two modules. With an Arduino Uno and drawbot/flutter for text printing, this device was created using the Raspberry Pi and a 1.2GHz processor. Doi: 10.28991/ESJ-2024-08-01-016 Full Text: PDF

Keywords

• text detection
• preprocessing
• segmentation
• character recognition
• process innovation
• product innovation.