Welcome to Learn photogrammetry in Finland blog
Photogrammetry is the art and science of using images, processing units, methods, and sensors to extract geometric and radiometric information, classification information, and relevant uncertainty about them. In photogrammetry, the main data are generally images that are classified into spatial, aerial, and short-range based on the distance to an object. Although images have a unique role in photogrammetry, other data such as point clouds, radar measurements, laser data, multi and hyperspectral images and a wide range of other data are also relevant.
Photogrammetry is closely linked to various scientific and technological disciplines, necessitating a foundational understanding of related subjects to grasp its concepts fully. For instance, knowledge of statistics and engineering probability is essential for accurately interpreting uncertainty in parameter estimates. Similarly, convex optimization techniques are employed to derive parameters, image processing is crucial for generating coded targets and point clouds, and neural networks assist in pattern recognition. Additionally, electronic principles are needed for measurement designs, among other relevant fields.
The current website, featuring selected parts of the book series Photogrammetric Computations (to be published next year), is the culmination of research and teaching in the fields of photogrammetry and computer vision. On this platform, I have aimed to simplify a concise list of traditionally difficult concepts for my students in the most accessible way possible. The content primarily reflects the photogrammetry calculations course that I introduced in a new format for undergraduate students in 2023. I decided to summarize and refine the teaching methods and materials I had developed, rewriting them to make them accessible to students worldwide.
In this blog, I provide essential background information in the simplest terms and gradually explain complex concepts. These are then implemented, at least partially, using programming languages like Python. Some of the methods presented here are experimental techniques I have created specifically for educational purposes. While an experimental method may not always be the most optimal approach for solving a problem, it enhances our understanding of the presented topics. This deeper comprehension is the primary goal of this blog.
Photogrammetry usually starts with images that taken with a digital camera from a small object like a pot, or a big object like a piece of land. Then images are linked together via 3D vision techniques, and consequently after few photogrammetric processing steps, 3D position of objects and cameras and uncertainties regarding those estimates are calculated:
Finally, a point cloud is calculated
Photogrammetry plays a crucial role in mobile mapping systems, where multiple sensors are integrated into vehicles or backpacks to capture extensive data from our streets. A notable example is the Google Street View car, which captures 360-degree images for Google Maps.
Today, photogrammetry is almost present wherever digital cameras are used for 3D measurement. This includes applications such as autonomous vehicles, robotic systems, airplanes, and advanced vision-based navigation systems, among others.
Throughout these projects, I am deeply grateful to my colleagues, whose contributions have been invaluable in helping me better understand various subjects, algorithms, and code. May their names be immortalized in the pursuit of advancing science and knowledge.
Ehsan Khoramshahi,
Helsinki, 2024