After adapting my algorithm dedicated to the extraction of digital terrain model (DTM) from terrestrial LiDAR point cloud, it is possible to process point clouds generated from imagery by photogrammetry software. Here is an example of DTM coming from drone data:
I developed a new Android app that collects air quality information online and presents different pollutants concentrations and the overall air cleanliness at the closest station. This side project has been done fully using the framework Ionic and is available for free on the Google Play Store The data comes from The World... Continue Reading →
This method finely approximates the ground surface by relying on deep learning to separate vegetation from potential ground points, files holes by blending multiple local approximations through partition of unity principle, then improves the accuracy by pushing the surface towards the data points through an iterative convection model.
Inspired by the EDL shading technique and SSAO, I implemented a new shader for OpenGL ES designed to enhance the rendering of point clouds. As the shading function rely only on the projected depth, and because the variable gl_FragDepth is missing in OpenGL ES, a workaround was to implement a custom depth buffer using a... Continue Reading →
Based on deep learning and a class decision process, I implemented a method designed to separate leaf points from wood points in terrestrial LiDAR point clouds of trees. Although simple, this approach learns trees characteristic point patterns efficiently and robustly. Comparison of wood / leaves segmentation methods using machine learning: https://youtu.be/sZ8PB1XHyeo
The action of the wind in the branches of cherry trees simulated and modeled with The grove and rendered in Blender https://youtu.be/bEDPpuiPeTg
A part of my PhD works is now available in the last release of the open source platform Computree. This wiki gives an overview of the computational steps implemented.
Departing from traditional tree models made of cylinders (in green), sharper models (in blue) can be computed. Expressed as a continuous surface, they describe the shape better and allow to assess precisely the tree properties. The 3D samples and the surface model of a tropical tree can be visualize here
Sequence of processing steps: modelling of the terrain surface, detection of the tree stem, then estimation of the tree model. 3D model of a Mangrove tree (Avicennia germinans Copyright A. Olagoke and C. Proisy, 2014, IRD UMR AMAP)