Photogrammetry 2 Explained with Simple Animations

Photogrammetry begins with capturing multiple photos of an object from different angles.The camera needs to move in a circular pattern around the subject, maintaining consistent distance.Each photo must overlap with the next by at least sixty percent. This ensures every surface detail is captured from multiple angles.Consistent lighting is crucial. The light source should remain fixed relative to the subject.This systematic approach creates a comprehensive dataset, ensuring complete coverage of the subject.As you move around the subject, maintain steady camera height and distance. Take photos at regular intervals to maintain consistent overlap.With our photos captured correctly, we're ready for the next step in the photogrammetry process.The software begins by analyzing each photo to identify distinct features.It looks for three main types of features: corners, edges, and distinctive texture patterns.For each detected feature, the software creates a mathematical descriptor that captures the unique characteristics of that point.These descriptors are like fingerprints, allowing the software to identify the same point across different images.The software then compares these descriptors across different images to find matching pairs.Let's look at the step-by-step process the algorithm follows.After detecting features, it calculates unique descriptors for each point.Then it matches features by comparing their descriptors across images.Finally, it filters out incorrect matches using geometric constraints.By analyzing these matched features, the software can determine the relative positions of the cameras when the photos were taken.This information about camera positions is crucial for the next step: reconstructing the three-dimensional structure.The 3D reconstruction process begins by triangulating the position of matched features in three-dimensional space.Using the known camera positions and matched feature points, the software calculates precise coordinates for each point, creating a point cloud.These points are then connected to form a triangular mesh. Each triangle is created by connecting three nearby points, gradually building the object's surface geometry.The final step involves mapping the original photo textures onto this geometric mesh. The software projects image data from the source photos onto each triangle of the mesh.The software continues to refine the model, adjusting vertex positions and optimizing texture mapping to create an accurate representation of the original object.The result is a detailed three-dimensional model that captures both the geometric shape and surface appearance of the original object.