登录

Photogrammetry-Guide

mikeroyal长期未更新

一份摄影测量入门与资源指南，梳理从无人机航拍、激光雷达到多视图立体的三维重建流程，覆盖测绘、建筑、农业、机器人和考古等应用场景。

Photogrammetry Guide. Photogrammetry is widely used for Aerial surveying, Agriculture, Architecture, 3D Games, Robotics, Archaeology, Construction, Emergency management, and Medical.

暂无版本1,536star110fork同步于 2026-11-07 07:33 UTC
摄影测量三维重建多视图立体激光雷达地理空间
对比 GitHub 官网3d-reconstructioncameracesiumdronesgeometry

怎么上手

README 里没有找到可直接复制的安装命令
原文

README 里没有找到可直接复制的安装命令，请到 README 或官方文档查看安装方式。

硬件要求按 README 说明

README 未说明

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼
没人工点评？写下你的经验

一份摄影测量学习指南，汇总应用场景、库与工具，帮工程师快速建立领域全貌

一份摄影测量入门与资源指南，梳理从无人机航拍、激光雷达到多视图立体的三维重建流程，覆盖测绘、建筑、农业、机器人和考古等应用场景。

核心能力

  • 梳理航空、近景、立体三类摄影测量方法
  • 讲解 SfM、SIFT、DEM/DSM/DTM 等核心概念
  • 覆盖点云处理、LiDAR、SLAM 与 NeRF 专题
  • 汇总摄影测量工具、库与框架及 Autodesk、游戏开发资源
  • 附课程、书籍与 YouTube 教程等学习材料

容易踩的坑

这是文档型指南仓库，不含可安装的软件或代码；README 未给出任何安装或运行命令；内容以概念和资源链接为主，需自行按链接获取具体工具

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

最近发版

0 个版本
全部 release

暂无 release 数据（可能没有发版，或尚未同步）。

用户点评

登录用户的真实使用记录，可按版本和硬件筛选
共 0 条

    还没有人点评。第一个说出真实体验的人最有价值。

    README 原文

    未翻译
    在 GitHub 打开

    Maintenance

    A guide covering Photogrammetry including the applications, libraries and tools that will make you a better and more efficient Photogrammetry development.

    Note: You can easily convert this markdown file to a PDF in VSCode using this handy extension Markdown PDF.

    Table of Contents

    1. Getting Started with Photogrammetry

    2. Photogrammetry Tools, Libraries, and Frameworks

    3. Autodesk Development

    4. LiDAR Development

    5. Game Development

    6. Machine Learning

    7. Python Development

    8. R Development

    Getting Started with Photogrammetry

    Back to the Top

    Photogrammetry is the art and science of extracting 3D information from photographs. The process involves taking overlapping photographs of an object, structure, or space, and converting them into 2D or 3D digital models. Photogrammetry is often used by surveyors, architects, engineers, and contractors to create topographic maps, meshes, point clouds, or drawings based on the real-world.

    Types of Photogrammetry

    • Aerial photogrammetry is process of utilizing aircrafts to produce aerial photography that can be turned into a 3D model or mapped digitally. Now, it is possible to do the same work with a drone.

    Image credit: Forestrypedia

    • Terrestrial (Close-range) photogrammetry is when images are captured using a handheld camera or with a camera mounted to a tripod. The output of this method is not to create topographic maps, but rather to make 3D models of a smaller object.

    Image credit: TLT Photography

    • Stereo Photogrammetry is a process that involves the estimation of 3D coordinates of points on an object by considering the measurements made of two or more images taken from different positions. The sensor looks at angles that are horizontal or oblique to collect accurate, detailed information for primarily engineering purposes, such as mapping the infrastructure of bridges, buildings, and dams.

    Image credit: Stereo Photogrammetry Capture. HELImetrex

    Photogrammetry Techniques

    Back to the Top

    • Structure from motion (SfM) is a photogrammetry range imaging technique for estimating 3D structures from 2D image sequences that may be coupled with local motion signals.

    • Scale-Invariant Feature Transform (SIFT) is a computer vision algorithm to detect, describe, and match local features in images. This includes object recognition, robotic mapping and navigation, image stitching, and 3D modeling.

    • Digital Outcrop Model (DOM) is a digital 3D representation of the outcrop surface, mostly in a form of textured polygon mesh. This allows for interpretation and reproducible measurement of different geological features, orientation of geological surfaces, width, and thickness of layers.

    • DEM (Digital Elevation Models) is a bare-earth raster grid referenced to a vertical datum. When you filter out non-ground points such as bridges and roads, you get a smooth digital elevation model. The built (power lines, buildings, and towers) and natural (trees and other types of vegetation) aren’t included in a DEM.

    • DSM (Digital Surface Models) is a process that captures both the natural and built/artificial features of the Earth’s surface.

    • DTM (Digital Terrain Models) is a process that has two definitions depending on where you live. USGS LiDAR Base Specification.

      • In some countries, a DTM is actually synonymous with a DEM. This means that a DTM is simply an elevation surface representing the bare earth referenced to a common vertical datum.

      • In the United States and other countries, a DTM has a slightly different meaning. A DTM is a vector data set composed of regularly spaced points and natural features such as ridges and breaklines. A DTM augments a DEM by including linear features of the bare-earth terrain.

    • TIN (Triangular Irregular Networks) is a representation of a continuous surface consisting entirely of triangular facets (a triangle mesh ), used mainly as Discrete Global Grid in primary elevation modeling. TINs are a form of vector-based digital geographic data and are constructed by triangulating a set of vertices (points). The vertices are connected with a series of edges to form a network of triangles. There are different methods of interpolation to form these triangles, such as Delaunay triangulation or distance ordering.

    • Canopy Height Model (CHM) is a separate model derived from elevation data in the point cloud. In forested areas, the difference between the DSM and the DEM can be viewed as CHM, representing the height of trees in the area above ground-level. Software utilizing CHMs can also derive individual tree data, such as crown diameter, crown area and tree boundaries.

    • Ground Control Points (GCPs) is a technique with known real-world coordinates that can be clearly identified in an image. GCPs are used in the orthorectification process to augment the geometric parameters embedded in the raw image and improve the accuracy of the resulting orthorectification.

    • Orthorectification is a process of removing distortion from an image caused by the curvature of the Earth and changes in terrain. It involves correcting the perspective of the image to align it with a map or a coordinate system, resulting in a more accurate representation of the Earth's surface.

    • Orthorectified imagery is a type of imagery that has been processed to apply corrections for optical distortions of satellite/aerial imagery caused by topography and sensor geometry errors.

    • Real Time Kinematic (RTK) is a satellite navigation technique used to enhance the precision of position data derived from satellite-based positioning systems (global navigation satellite systems, GNSS) such as GPS, GLONASS, Galileo, NavIC and BeiDou.

    • Post Processing Kinematic (PPK) is an alternative a satellite navigation technique to Real-Time Kinematic (RTK) used in surveying that corrects the location data after it is collected and uploaded. This technique is commonly used in drone mapping and land surveying.

    • Gaussian Splatting is a rasterization technique described in 3D Gaussian Splatting for Real-Time Radiance Field Rendering that allows real-time rendering of photorealistic scenes learned from small samples of images. Polycam Gaussian splat viewer and creator

    • DynIBaR (Neural Dynamic Image-Based Rendering) is a new method that generates photorealistic free-viewpoint renderings from a single video of a complex, dynamic scene. It can be used to generate a range of video effects, such as “bullet time” effects (where time is paused and the camera is moved at a normal speed around a scene), video stabilization, depth of field, and slow motion, from a single video taken with a phone’s camera. DynIBaR: Space-time view synthesis from videos of dynamic scenes | Google Research

    Cameras

    Back to the Top

    Links to Helpful Resources:

    General camera-type recommendations

    Here are some categories in which you should start your research:

    Fully automatic operation:

    • Default: high-end phone.

    • Some more flexibility in a zoom range: high-end compact camera (as in the Sony RX100 line).

    • More flexibility, including distant subjects: high-end superzoom camera (as in the Sony RX10 and Panasonic FZ1000 lines).

    • Extremely distant subjects in daylight: consumer superzoom camera (as in the Nikon P line, Panasonic FZ80, etc.).

    Manual/creative control:

    • Default: mirrorless camera.

    • Cheaper alternative (no mirrorless camera that suits your needs is available under budget): DSLR.

      If you’re absolutely sure that’s the only lens you’ll need: fixed prime lens camera (as in the Fujifilm X100 and Ricoh GR lines), high-end compact camera or high-end superzoom camera.

    Key concepts and terminology

    There are some concepts, terms and features of a camera that you’ll need to learn about to really understand camera reviews and see how one camera differs from another. The following is a list of such terms for you to look up if needed.

    • Exposure, noise, dynamic range (SDR and HDR).

    • Camera design: interchangeable-lens cameras and fixed-lens cameras, mirrorless and DSLR.

    • Image sensor: size and surface area, resolution.

    • Lens: focal length and angle/field of view, maximum aperture, lens mount and format coverage.

    • Autofocus.

    • Continuous/burst shooting, buffer depth.

    • Viewfinder and display.

    • Image stabilization.

    • Weather resistance.

    Remember a camera’s age is irrelevant. Cameras don’t age like smart phones or computers do, because they have no increasingly demanding software to keep up with. So as long as a camera is in good working order, it should work as well as it did when it was brand new.

    Make sure to shop at well accepted and well established in the camera market, even at the high end. Try reputable outlets (KEH and mpb) and the used sections on big retailers (B&H and Adorama in the US) and local camera stores. You can also find refurbished cameras sold directly by the manufacturers’ distributors.

    Drones

    Back to the Top

    Links for Drone photography:

    A drone is basically an Unmanned Aerial Vehicle (UAV). Before the rise in consumer interest in UAVs, the word “drone” was primarily used to refer to the UAVs used by the military.

    Now, though, intelligent quadcopters that have UAV-like features are more popular among consumers than ever before. And while they technically aren’t as advanced as military drones, we refer to them as ”drones” because they are similar in nature (both allow you to operate an aerial vehicle in order to perform a particular task, which, in the case of consumer drones, is typically to shoot video or capture still images).

    There are different types of drones for all kinds of situations, so make sure to consider the main purpose for your drone purchase. Even though most drones fall under the categories of consumer or professional grade drones, there are now UAVs geared towards traveling and selfies.

    • Camera drones: These drones might range from $100 to $1000 which have exceptional features that make such drones a market leader in the world of professional drones. They also have Obstacle avoidable sensors.

    • Compact drones: These drones are specially designed for panoramic images and landscape shoot. The best feature of compact drones is that they have an ability to change the image format from JEPG to raw images. Raw images have more details and can produce 4k quality of the picture and videos. For example, DJI Inspire 2, which falls in the category of professional drones has a speed of 60+ miles per hour (97+ kilometers per hour) and has a dual battery backup of up to 30 minutes. Use these drones in the places where you cannot step in and take the pictures that you don’t need to crop or it can adversely affect the quality of the picture.

    The different among RTF, BNF and ARF:

    • RTF stands for Ready-To-Fly - Usually an RTF quadcopter doesn’t require any assembly or setup, but you may have to do some simple things like charge up the battery, install the propellers or bind the controller to the quadcopter (get them talking to each other).

    • BNF stands for Bind-And-Fly - A BNF quadcopter usually comes completely assembled, but without a controller. With BNF models, you’ll have to use the controller that you already have (if it’s compatible) or find a controller sold separately. One thing you should know is that just because a transmitter and receiver are on the same frequency that doesn’t mean that they’ll work together.

    • ARF stands for Almost-ready-to-fly - ARF Drones are usually like quadcopter kits. They usually don’t come with a transmitter or receiver and might require partial assembly. An ARF drone kit might also leave out components like motors, ESCs, or even the flight controller and battery. The definition of an ARF drone kit is very broad, so whenever you see ARF in the title, you should read the description thoroughly.

    Taking care of legal consideration

    Before you get started with your drone (I know you must be excited) make sure you are updated with all the rules and regulations for what you can and what you cannot do with the drone, the restricted areas and the permission to fly near any private property.

    • Various countries have their own rules and regulations for flying drones. For Example in the UK it’s Civil Aviation Authority (CAA) that restricts you from flying drones near airports or aircraft, keep the drone below 400 feet etc. Further you can refer to the UK link

    • In the US it’s federal Aviation Administration (FAA) which has a code of not to fly above a group of people, never fly under influence of drugs and alcohol. Please refer to here

    • Similarly, for various countries you have strict rules that need to be taken care of before making a decision of buying a drone. So, save yourself from getting disheartened of the fact that you cannot achieve the purpose for what you bought this drone.

    Improve the drone flying skills

    A lot of people think that drones are hard to fly, but the truth is, they’re really not. Anyone capable of using an iPhone or Android device is more than capable of flying a drone. However, this does not mean that drones are fool proof. Even the most advanced drones require some general knowledge if you want to avoid crashing or worse, losing your drone forever. So you need to improve your flying drone skills.

    Where To Buy A Drone?

    If you don’t know where to buy a drone, don’t worry. There are tons of online stores for drones that will ship to just about any major country. If you’re buying toy drones, the best place to go is Amazon or others below:

    All of the main websites for buying drones.

    Geographic Information System

    Back to the Top

    Geographic Information System (GIS) is an information system able to encode, store, transform, analyze and display geospatial information.

    Image credit: geo.university

    Geographic Information System Software

    • ArcGIS Desktop: Extendable desktop suite to manage, visualize and analyze GIS data in 2D and 3D, including image processing. Includes ArcGIS Pro, ArcMap, ArcCatalog, and ArcGIS Online.
    • DIVA-GIS - DIVA-GIS is a free geographic information system software program used for the analysis of geographic data, in particular point data on biodiversity.
    • GeoDa - A free and open source software tool that serves as an introduction to spatial data analysis.
    • GISInternals - Povidesdaily build packages and software development kits for the GDAL and MapServer
    • Global Mapper - An easy-to-use, robust, and genuinely affordable GIS application that combines a wide array of spatial data processing tools with access to an unparalleled variety of data formats.
    • GRASS GIS - A free and open source GIS software suite used for geospatial data management and analysis, image processing, graphics and maps production, spatial modeling, and visualization.
    • gvSIG - A powerful, user-friendly, interoperable GIS.
    • JUMP GIS - An open source GIS written in Java
    • MapInfo Pro - A full-featured desktop solution to prepare data for web mapping applications and create presentation quality maps that combines data analysis, visual insights, and map publishing.
    • Marble - A virtual globe and world atlas.
    • OpenOrienteering Mapper - A software for creating maps for the orienteering sport.
    • QGIS - A free and open source GIS.
    • SAGA - Open source system for automated geoscientific analyses.
    • SharpMap - An easy-to-use mapping library for use in web and desktop applications
    • TileMill - An open source map design studio, developed by a community of volunteer open source contributors
    • Whitebox GAT - An open source desktop GIS and remote sensing software package for general applications of geospatial analysis and data visualization.
    • DIVA-GIS - DIVA-GIS is a free geographic information system software program used for the analysis of geographic data, in particular point data on biodiversity.
    • Abc-Map - A lightweight and user-friendly Web GIS. Create, import data from various sources, export maps or share them online freely and easily.

    Remote Sensing

    Back to the Top

    Remote Sensing is a set of techniques used to gather and process information about an object without direct physical contact.

    • Active remote sensing are instruments that operate with their own source of emission or light, while passive ones rely on the reflected one. Radiation also differs by wavelengths that fall into short (visible, NIR, MIR) and long (microwave). Active remote sensing techniques differ by what they transmit (light or waves) and what they determine (distance, height, atmospheric conditions, etc.).

    • Passive remote sensing are instruments that depend on natural energy (sunrays) bounced by the target. For this reason, it can be applied only with proper sunlight, otherwise there will be nothing to reflect. It employs multispectral or hyperspectral sensors that measure the acquired quantity with multiple band combinations. These combinations differ by the number of channels (two wavelengths and more). The scope of bands (visible, IR, NIR, TIR, microwave).

    Image credit: mdpi

    Remote Sensing Software

    • eCognition - A powerful development environment for object-based image analysis.
    • ENVI :star2: - A geospatial imagery analysis and processing software.
    • ERDAS IMAGINE :star2: - A geospatial imagery analysis and processing software.
    • Google Earth - A computer program that renders a 3D representation of Earth based on satellite imagery.
    • Google Earth Studio - An animation tool for Google Earth’s satellite and 3D imagery.
    • GRASS GIS - A free and open source GIS software suite used for geospatial data management and analysis, image processing, graphics and maps production, spatial modeling, and visualization.
    • Opticks - An expandable remote sensing and imagery analysis software platform that is free and open source.
    • Orfeo toolbox - An open-source project for state-of-the-art remote sensing, including a fast image viewer, apps callable from Bash, Python or QGIS, and a powerful C++ API.
    • PANOPLY- Panoply plots geo-referenced and other arrays from netCDF, HDF, GRIB, and other datasets.
    • PCI Geomatica - A remote sensing desktop software package for processing earth observation data.
    • SNAP - A common architecture for all Sentinel Toolboxes.

    Point Cloud Processing

    Back to the Top

    Point Cloud Processing is a huge number of tiny data points that exist in three dimensions(3D). If you could spit those points out of a scanner they would appear as a cloud. This can be used for mapping, perception, and navigation in robotics and autonomous systems. It can also be used in augmented reality (AR) and virtual reality (VR) applications.

    Functions of point cloud processing software include:

    • Manual editing of the point cloud.
    • Automatic cleanup and filtering.
    • Offloading tasks to the cloud.
    • Rendering.
    • Point distance calculation.
    • Registration.
    • Conversion to mesh.
    • Conversation to NURBS.

    Point Cloud Processing & Data Management. Image credit: H2H Associates

    Point Cloud Processing Software

    • CloudCompare is a 3D point cloud processing software. It can also handle triangular meshes and calibrated images.
    • PCL - Point Cloud Library is a standalone, large scale, open project for 2D/3D image and point cloud processing.
    • Leica Cyclone is the market-leading point cloud processing software. It is a family of software modules that provides the widest set of work process options for 3D laser scanning projects in engineering, surveying, construction and related applications.
    • FARO SCENE is a software that allows you to capture, process and register real-world objects and environments in 3D point cloud. You can create stunning visualizations, share and collaborate, and access virtual reality -LRB- VR -RRB- views of your scans.
    • Autodesk ReCap is a software that let's you create accurate 3D models with reality capture using laser scanning. This software, which lets you convert reality into a 3D model or 2D drawing, actually comes in two flavors: ReCap and ReCap Pro.
    • Trimble RealWorks is a powerful office software suite to integrate 3D point cloud and survey data. It effectively manages large scan datasets to register, analyze, model, collaborate and produce compelling deliverables.
    • PDAL - Point Data Abstraction Library is a C++/Python BSD library for translating and manipulating point cloud data.
    • libLAS is a C/C++ library for reading and writing the very common LAS LiDAR format (Legacy. Replaced by PDAL).
    • entwine is a data organization library for massive point clouds, designed to conquer datasets of hundreds of billions of points as well as desktop-scale point clouds.
    • PotreeConverter is another data organisation library, generating data for use in the Potree web viewer.
    • lidR is a R package for Airborne LiDAR Data Manipulation and Visualization for Forestry Applications.
    • pypcd is a Python module to read and write point clouds stored in the PCD file format, used by the Point Cloud Library.
    • Open3D is an open-source library that supports rapid development of software that deals with 3D data. It has Python and C++ frontends.
    • cilantro is a Lean and Efficient Library for Point Cloud Data Processing (C++).
    • PyVista is a 3D plotting and mesh analysis through a streamlined interface for the Visualization Toolkit(VTK).
    • pyntcloud is a Python library for working with 3D point clouds.
    • pylas Reading Las (lidar) in Python.
    • PyTorch PyTorch Geometric (PyG) is a geometric deep learning extension library for PyTorch.
    • Paraview is a Open-source, multi-platform data analysis and visualization application.
    • MeshLab is a Open source, portable, and extensible system for the processing and editing of unstructured 3D triangular meshes
    • OpenFlipper is a Open Source Geometry Processing and Rendering Framework.
    • PotreeDesktop is a desktop/portable version of the web-based point cloud viewer Potree
    • 3d-annotation-tool is a lightweight desktop application to annotate pointclouds for machine learning.

    LiDAR

    Back to the Top

    Light Detection and Ranging (LiDAR) is a technology used to create high-resolution models of ground elevation with a vertical accuracy of 10 centimeters (4 inches). Lidar equipment, which includes a laser scanner, a Global Positioning System (GPS), and an Inertial Navigation System (INS), is typically mounted on a small aircraft. The laser scanner transmits brief pulses of light to the ground surface. Those pulses are reflected or scattered back and their travel time is used to calculate the distance between the laser scanner and the ground.  Lidar data is initially collected as a “point cloud” of individual points reflected from everything on the surface, including structures and vegetation. To produce a “bare earth” Digital Elevation Model (DEM), structures and vegetation are stripped away.

    3D Data Visualization of Golden Gate Bridge. Source: USGS

    Mola is a Modular Optimization framework for Localization and mApping (MOLA).

    3D LiDAR SLAM from KITTI dataset. Source: MOLA

    Basic matching algorithms

    Semantic segmentation

    Ground segmentation

    Simultaneous localization and mapping SLAM and LIDAR-based odometry and or mapping LOAM

    Object detection and object tracking

    Neural Radiance Field (NeRF)

    Back to the Top

    NeRF (Neural Radiance Field) is a neural rendering model that learns a high-resolution 3D scene in seconds and can render images of that scene in a few milliseconds. It works by taking a handful a of 2D images representing a scene and interpolating between them to render one complete 3D scene.

    NVIDIA Instant NeRFs. Image Credit: NVIDIA

    Tiny CUDA Neural Networks is a small, self-contained framework for training and querying neural networks. Most notably, it contains a lightning fast "fully fused" multi-layer perceptron (technical paper), a versatile multiresolution hash encoding (technical paper), as well as support for various other input encodings, losses, and optimizers.

    NVIDIA Instant NeRFs using tiny-cuda-nn (Tiny CUDA Neural Networks). Image Credit: NVIDIA

    LERF(Language Embedded Radiance Fields)

    Back to the Top

    LERF(Language Embedded Radiance Fields) is a AI process that optimizes a dense, multi-scale language 3D field by volume rendering CLIP embeddings along training rays, supervising these embeddings with multi-scale CLIP features across multi-view training images.

    LERF Rendering. Image credit: LERF.io

    Certifications & Courses

    Back to the Top

    Books/eBooks

    Back to the Top

    YouTube Tutorials

    Back to the Top

    Introduction to Photogrammetry (Cyrill Stachniss, 2021 [![Complete Meshroom Tutorial | Photogrammetry Course by Gleb Alexandrov](https://ytcards.demolab.com/?id=j3lhPKF8qjU&lang=en&background_color=%230d1117&title_color=%23ffffff&stats_color=%23dedede&width=240 "Complete Meshroom Tutorial | Photogrammetry Course by Gleb Alexandr

    相关项目

    同分类里 star 最高的项目
    对比全部 →
    gaussian-splatting
    graphdeco-inria·Python

    3D Gaussian Splatting 的官方参考实现，用高斯点云做辐射场重建，实现实时可交互的逼真场景渲染。

    3D高斯泼溅+43D 重建
    24449+173.5k2025-10-17 14:28 UTC
    instant-ngp
    NVlabs·Cuda

    用多分辨率哈希编码把 NeRF、SDF 等神经图形基元的训练从几小时压到几秒，实时渲染三维场景。

    神经辐射场+43D 重建
    17590+22.1kv2.02026-02-02 12:32 UTC
    Open3D
    isl-org·C++

    Open3D 是一个现代 3D 数据处理库，提供点云、网格、RGB-D 和 3D 可视化等算法与工具，支持 C++ 和 Python，可用于 3D 感知、重建与机器学习研究。

    3D数据处理+43D 重建
    14189+82.6kv0.20.02026-11-05 18:33 UTC
    colmap
    colmap·C++

    从多张照片重建三维场景的 C++ 工具，集成了运动恢复结构（SfM）和多视图立体（MVS），可输出相机位姿与稠密点云。

    三维重建+43D 重建
    13189+152.2kdatasets-eth3d-undistorted-v12026-11-10 01:17 UTC