GeoAI-Based Nonpoint Pollution Source Management System
There are plenty of images, but not enough people to analyze them... and organizing the results is even harder.
Administrative reality: field surveys are difficult, but organizing and analyzing the data is even more challenging.
Growing project demands for attribute information, reports, and quantitative analysis.
Increasing need for a repeatable analysis framework.
Heavy burden of manpower input.
Therefore, an automated analysis system is essential.
It is designed to automate complex analysis and reporting tasks, enabling anyone to easily perform everything from detection to report generation.
Web Service (Optimized for Visual Communication)
- Supports map-based visualization of detection results
- Integrated analysis by individual image unit
- Summary of the number and area of pollution sources such as compost piles, livestock sheds, and agricultural waste
- Generation of graph-based quantitative reports
GUI Program (Optimized for Field Application and Practical Work)
- Allows direct specification of orthophotos and storage paths
- Detection reports can be viewed immediately (offline execution possible)
- Supports checking detailed attribute information for individual nonpoint pollution sources
- Optional mailing service functionality
- Automatically generates summary statistical reports by municipality
A Practical, Data-Driven Solution for Water System Pollution Management
Nonpoint pollution sources must transition from being “visually observed” to being “data-managed.”
IREM Technology Development Co., Ltd.’s compost pile detection system goes beyond simply identifying pollution sources — it calculates area, confirms location, determines classification levels, and generates comprehensive reports.
From investigation to reporting, this automated solution significantly reduces the field survey burden for local governments and institutions while dramatically improving management efficiency.
1
A Practical, Data-Driven Solution for Water System Pollution Management
Nonpoint pollution sources must transition from being “visually observed” to being “data-managed.”
IREM Technology Development Co., Ltd.’s compost pile detection system goes beyond simply identifying pollution sources — it calculates area, confirms location, determines classification levels, and generates comprehensive reports.
From investigation to reporting, this automated solution significantly reduces the field survey burden for local governments and institutions while dramatically improving management efficiency.
2
Operational Flexibility
Supports both standalone GUI and web-based systems
User-customizable settings such as class-based and distance-based criteria
Enables both individual detailed analysis and batch analysis
3
Practical Utilization of Generated Results
Immediate result export in formats such as CSV and SHP
Automatic application of report templates
Provides outputs compatible with GIS programs for integrated analysis
System Application Cases
- (2020–2022) Nakdong River Watershed Management Committee
“Detection of Water System Pollution Sources and Water Environment Management Using Drones and Artificial Intelligence”
- (2023, 2024) Geum River Watershed Management Committee
“Study on Water System Pollution Source Investigation and Water Environment Management Using Drones”
- (2022, 2024) Nakdong River Basin Environmental Office
“Environmental Aerial Monitoring Operations Using Drones”
Possesses practical application experience primarily with national and local environmental agencies!
System Application Detection Cases
CASE 1 Compost Piles (Covered or Uncovered)
CASE 2 Compost Piles (Covered or Uncovered) / Livestock Sheds / Mulching Film / Greenhouses
1
Enables rapid and precise detection of multiple pollution sources without manual work
2
Flexible response by detecting only the pollution source types selected by the user
3
Detection confidence for all objects is 0.8 or higher, ensuring high accuracy and suitability for field application
4
Accurately detects targets that are difficult to identify visually, such as abandoned livestock manure in sheds or mixed pollution sources in bare land
5
Deep learning-based algorithms automate numerical information such as area, coordinates, and volume without human subjectivity
Pollution Source Analysis Process
1. Drone Filming
Conduct exploratory flights and aerial imaging for the management of nonpoint pollution sources in agricultural areas.
- Set flight paths and imaging sections for target areas such as riversides and farmland
- Maintain a consistent altitude to capture high-resolution images
- Aerial reconnaissance for initial exploration of pollution source distribution
※ Integrated operation, including drone filming services, can be provided based on the needs of the implementing agency
2. Image Preprocessing
Generate high-precision analysis data from high-resolution drone imagery.
- Automatically organize and correct captured image data using image processing software
- Produce distortion-free orthomosaic images and DSM (Digital Surface Models)
- Secure spatial reference data for accurate numerical analysis of area, volume, and other measurements
- Possess in-house expertise for the entire process, including imaging and preprocessing
3. AI-Based Detection and Analysis
Automatically identify major pollution sources and calculate attributes using object detection algorithms.
- Automatically detect user-specified objects among nonpoint pollution sources in agricultural areas using deep learning models
- Users can specify combinations of classes (e.g., compost piles, agricultural waste, livestock sheds, greenhouses)
- Automatically calculate attributes for detected objects, including center coordinates, area, volume, and distance to water systems
※ Optional use of either a standalone GUI program or a web-based system
4. Result Generation and Review
Provide analysis results that can be applied in the field and support decision-making.
- Automatically generate PDF reports, CSV attribute data, and SHP spatial information files after analysis
- Immediately usable in GIS programs such as QGIS and ArcGIS
- When linked to a database, results can be used for time-series comparisons and pre/post monitoring
※ Output formats and report templates can be customized to user requirements
