To improve recognition accuracy, the following technologies enhance image preprocessing capabilities:

• Trapezoidal Correction: Trapezoidal distortion may occur in images when photographing or scanning documents (for example, when photographing a document from an angle). Using a trapezoidal correction algorithm, images are automatically adjusted to a correct rectangle, further enhancing the effectiveness of OCR recognition while maintaining the accuracy of text structure.
• Automatic Cropping : The edges of documents are automatically detected and unnecessary background areas are cropped. This makes it easier to focus on the text content, minimizes distracting background elements, and improves the accuracy of OCR recognition results.
• Image Correction : High-precision geometric correction techniques are used to remove image distortion and tilt, restoring the original structure and visual appearance.

• We provide a RESTful API for external applications to access the recognition service, making it easy to integrate OCR recognition into existing workflows.
• Rapid Integration: Simply pass in the image and DocID when calling the API, and receive recognition results in JSON format.
• Real-Time Service: External applications can submit documents through the API and receive OCR results in real time. This enables fast processing of large document volumes and and enhances enterprise automation efficiency.

• We offer customizable document recognition settings, allowing users to add document types to recognize, define specific fields to recognize, and run real-time tests. If the results meet expectations, the settings can be saved for future document uploads.
• During the recognition and extraction process, the system automatically identifies fields with semantically similar labels (e.g., Date of Birth, Birth Date, Birth Year and Month, etc.). Therefore, uploading sample training data is not required to complete the setup.