Photogrammetry, derived from the Greek words "photo" meaning light and "metry" meaning measurement, is a scientific discipline that delves into the estimation of geometric and semantic properties of objects using photographs. The photographs can be acquired from:
By utilizing multiple photographs captured from different angles, photogrammetry enables the creation of both two-dimensional orthophotos and intricate three-dimensional models, offering a comprehensive representation of the scene. This versatility makes it ideal for change monitoring, urban growth assessment, natural disaster evaluation, and tracking dynamic processes. Moreover, photogrammetry has become more accessible with the advent of consumer-grade drones and advancements in camera technology, making it a cost-effective solution.
At its core, photogrammetry encompasses several key aspects that empower us to unlock the hidden secrets captured within images. By analyzing specific cues within the photographs, such as recognizable features or patterns, we can infer the camera's location and orientation relative to the scene it captured. This triangulation of multiple camera locations enables us to unravel the three-dimensional geometry of the scene, estimate the spatial positions of objects, and reconstruct their shapes with remarkable precision.
To accurately estimate the geometry of the scene from images, it is important to understand image acquisition concept particularly how points in the 3D world are mapped to the 2D images.
We have world/object coordinate system (X,Y,Z). A point in 3D world is represented by X, Y and Z.
We have camera coordinate system. Everything is defined with respect to the camera. [Xk, Yk, Zk].The mapping between the object coordinate system and the camera coordinate system is typically a rigid body transformation which is simply a shift in 3-dimention then rotation. Then we have the image (plane) coordinate system [xc, yc]. This is when we are going from the 3D to 2D.
The transformation equation from 3D to 2D is shown by
[xc,yc,1] = Hc Hk [X,Y,Z,1]
Extrinsic parameters describe the pose of the camera while Intrinsic parameters are internal or special to the camera and doesn’t change rapidly. Extrinsic parameters are 6: 3 positional parameters and 3 heading parameters.
Knowing the rotation R and the camera position Xo, jja point Xp in the world coordinate system can be mapped to the camera coordinate system using:
Xc = R(Xp- Xo)
The camera coordinate can then be mapped to the image plane.