Here's a draft for your README file based on the provided structure and additional information:

---

3D Reconstruction of Additive Manufactured Workpieces for Quality Control: Integrating Photogrammetry, Neural Radiance Fields, and Deep Learning

Abstract from this work's Paper

This thesis presents the development of a novel quality control tool for additive manufacturing workpieces, leveraging advanced 3D reconstruction and machine vision techniques. The sys- tem integrates state-of-the-art methods, including photogrammetry, Neural Radiance Fields (NeRF), and deep learning, to achieve precise reconstructions and automated defect detec- tion. The proposed pipeline enables high-speed 3D reconstruction with sub-minute execution times, delivering superior geometric accuracy compared to traditional photogrammetric ap- proaches. Validation is conducted using a custom-developed dataset including diverse object types, ensuring robustness across varying geometries and material properties. Key perfor- mance metrics include dimension recovery accuracy and average Chamfer and Hausdorff distances, offering comprehensive measures of reconstruction quality. Notably, the developed pipeline reduces the average Chamfer distance by 226% and the average Hausdorff distance by an impressive 1590% (lower values indicate better performance), while achieving total reconstruction times of just a few minutes and significantly outperforming traditional pho- togrammetry, which averages 31.16 minutes per reconstruction. Additionally, a quality con- trol application is developed, featuring a user-friendly interface for visualization, comparison, and defect analysis, making it highly applicable for industrial use. This work demonstrates the potential of combining cutting-edge 3D reconstruction techniques with deep learning to optimize and automate quality assurance processes.

---

Project Folders Description

This repository is organized into several key folders:

• lib: Includes utility functions for:
  • Calibration: Functions for handling Charuco board calibration.
  • JSON Generation: Scripts for creating and manipulating JSON files for camera poses and other metadata.
  • Geometry Functions: Helper functions for 3D geometric calculations.
• scripts: Python scripts called by the Node-RED server to perform tasks such as:
  • Image acquisition.
  • Pose estimation.
  • Mesh extraction and comparison.
  • Initiating reconstruction scripts.
• node-red: The main application folder. Contains Node-RED flows and configuration files. The application serves as a user interface for data visualization and quality control.
• environment.yml: Environment configuration file for setting up the required Python dependencies.

---

Parameter Description

The following are key parameters used in the project:

1. Charuco Board:

   • ARUCO_DICT: cv2.aruco.DICT_6X6_250 (Defines the marker dictionary used for calibration).
   • SQUARES_VERTICALLY: 13.
   • SQUARES_HORIZONTALLY: 9.
   • SQUARE_LENGTH: 0.03m.
   • MARKER_LENGTH: 0.02m.

2. Marching Cubes Parameters:

   • mc_resolution: Defines the resolution of the grid for mesh extraction. Higher values produce finer meshes.
   • mc_threshold: Threshold for determining surface boundaries in the scalar field.

3. Application Parameters:

   • Default Node-RED server port: 9800.
   • Path to save processed results and intermediate files.

---

Installation

To set up the environment and install required dependencies:

1. Create the Python Environment:

   conda env create -f environment.yml
   conda activate <environment-name>

2. Install Instant-NGP: Follow the instructions provided in the official repository: instant-ngp.
3. Install Colmap: Download and install Colmap from Colmap's official website.
4. Install Node-RED: Follow the local installation instructions from Node-RED's website.

---

Example of Usage

The workflow for obtaining and processing data:

1. Prepare the Environment:

   • Activate the environment: conda activate <environment-name>.
   • Start the Node-RED server:

     cd node-red
     node-red

   • Access the application via your web browser at http://localhost:9800.

2. Data Acquisition:

   • Use the interface to capture images with the connected camera.
   • Ensure the calibration files are ready for pose estimation.

3. Pose Estimation:

   • Use Colmap or Charuco tab for generating camera poses.

4. Mesh Extraction:

   • Run the reconstruction using API for instant-ngp algorithms with specified parameters.

---

Data

The validation dataset used in this project can be downloaded here. The dataset contains images, calibration files, and ground truth meshes for various test objects.

---

Important Remarks

1. Best Practices:

   • Always verify that the camera is calibrated before acquiring data. Misalignment in poses can result in inaccurate reconstructions.
   • Use consistent lighting conditions to ensure high-quality images for reconstruction.

2. Common Issues:

   • Node-RED Server Not Starting: Ensure the correct environment is activated and Node-RED is installed properly.
   • Pose Estimation Errors: Verify that the Charuco board is correctly placed in the scene and well-lit during image capture.
   • Mesh Extraction Artifacts: Fine-tune the mc_resolution and mc_threshold parameters to improve mesh quality.

---