GIS - Automatic tree recognition from orthophotos

Project Details

We have developed an innovative solution to automate the forest inventory process. Using modern geoinformation technologies and machine learning methods, we have created a system that allows us to accurately and quickly determine the location of trees over large areas. This greatly simplifies and speeds up the process of collecting data on forest resources.

GIS – Automatic tree recognition from orthophotos

Client: German Federal Forestry Administration

Sector:

Geographic information systems (GIS)
Machine learning
Environmental monitoring

Sub Sector:

Analyzing spatial data
Automatic object recognition
Algorithm development
Environmental research

Services:

LiDAR mapping
Analyzing geospatial data
Automatic object recognition
Digital map creation

Solution:

Automatic tree recognition from orthophotos.
Using machine learning algorithms for data analysis.
Integrating algorithms into GIS applications.

Location: Germany

Technology and Software:
Our project is based on the use of advanced geoinformation technologies. Thanks to the use of ArcGIS Pro and deep learning tools, we have achieved high accuracy in automatically locating trees. This allows us to obtain detailed maps of forest areas and use them for effective forest management.

People:
GisPoint performed the following tasks: The project manager was responsible for the overall planning, coordination and control of the project. GIS specialists collected data, developed algorithms, tested and implemented machine learning models in ArcGIS Pro. The quality controller assessed the accuracy of the results and ensured compliance with quality standards.

Process & Challenges

Process:
The project was implemented using the integrated ArcGIS Pro platform, which ensured efficient data and computing management. The implementation process included the following stages:

Development of the classification algorithm: Using ArcGIS Pro software and machine learning tools, an algorithm was developed to automatically classify image pixels and locate each tree.
Model testing and validation: The accuracy of the developed model was tested on an independent sample of data compared to manual counting and postal surveys.
Map generation: Based on the classification results, detailed digital maps of each tree location were created.

Challenges:
Automatically finding tree locations with ArcGIS Pro is a complex process that comes with a number of challenges. Here are some of the most common issues we’ve encountered:

Noise and artifacts: Shadows, clouds, atmospheric phenomena, as well as various types of noise can interfere with segmentation algorithms.

Spectral characteristics: The diversity of tree species, their condition (healthy, diseased, young, old), and weather conditions can lead to significant variations in spectral characteristics, making it difficult to classify them.

Computing resources: Processing large amounts of data and complex machine learning models required powerful hardware.

Training sample preparation: Creating a high-quality training set with accurate annotations was a labour-intensive process.

These challenges, as we know, are an integral part of such projects and required us to carefully analyze and find optimal solutions.

Result

The project resulted in a high accuracy of tree coordinates (90%), which allows us to obtain detailed information about their location in the forests. A significant amount of data (600 hectares) was processed and detailed tree distribution maps were created.

Our Team

Ievgen Lavrishko

CEO & Owner

Khrystyna Bochko

HR Generalist

Alevtyna Kostianchuk

Senior Project Manager

Andriiana Pavlyshyn

Head of Business Development

Roman Nahaiovskyi

Project Manager

Maria Kizim

Lead Generation Specialist

Ivanna Soltys

Tender Manager

Navigating Our Impressive GIS Portfolio

GIS Analysis for Determining Optimal Locations for Wind Farms

Location:

Australia

As part of the project, we developed KML files to depict various constraints within the study area. Moreover, we produced wind speed images using data from the global wind atlas.
Digitizing utility service connections

Location:

Canada

The project encompasses the digitization of utility service connections, specifically for water, sewer, and storm services, spanning across 12,706 properties (parcels) within the City.
Fulfillment of government orders to create and update databases

Location:

Poland

The databases are created and updated by vectorizing existing raster data (topographic maps), orthophotomaps and processing geodetic files and land documentation (by coordinates and sketches).
Geospatial Analysis and Image Processing

Location:

Poland

Editing the automatically generated database of motorways, bridges and roads.
Precision 2D Vector Map Creating at 1500 Scale

Location:

Latvia

Precision 2D Vector Map Creating at 1500 Scale