I enjoy taking photos of interesting subjects whenever I get the opportunity. I use these photos to create 3D models through photogrammetry. One of the subjects that I photographed multiple times was an old steam train located on the west side of Tehran’s Mellat Park. The train was placed there for children to play with. If I recall correctly, I visited this section of the park multiple times last year and captured pictures each time. At last, I took around 2400 photos of different parts of the train using my Nokia 5.3 mobile phone.
One of the challenges I faced was finding the right time to take photos since this area of the park receives direct sunlight most of the time. However, I managed to take some photos during cloudy weather with soft shadows and some early in the morning before sunrise. However, I had to take some photos in direct sunlight. The next issue I encountered was the movement and repositioning of parts that could be adjusted, such as the front and back chains and some handles that were attached to the train by children who were playing with them. This eventually affected the final result.
Because aligning all the photos did not produce the right result, I divided the photos into four groups: the top of the train, the right side, the front and left sides together, and inside the cab, and I aligned them separately as displayed in the images below:
Now the time-consuming step of matching the components using control points. First, the inside of the cab matched with the component of the top part, and a new component was created as displayed in the image below:
Then, the result of matching the two components of the top and the inside of the cab was matched with other components. Finally, eleven groups of control points were set to match all the components, the result of which is shown in the image below:
In the next step, I used the MeshLab software to edit the points to remove the extra points and make some other edits. Then I thought uploading a cloud points version to the Sketchfab website might be a good idea.
In the video below, you can see the first 3D model with about 10 million triangle faces. Although the result seems to be good in general, some parts need to be corrected. Perhaps it would be better to use the retopology technique to modify the model.
This blog post will be updated in the future.
If you have any questions or comments, feel free to contact me or leave them in the comments section below.
