Principal investigators: Song Zhang, professor, mechanical engineering, Purdue University (zhan2053@purdue.edu); Songlin Fei, professor, Forestry and Natural Resources, Purdue University.
Co-authors: Victor Chen, Cheryl Qian, Rado Gazo, Wang Xiang, Zhiheng Yin, Zizun Zhou
Forest health and sustainability rely on accurate and efficient monitoring, with tree inventories playing a critical role. The initial goal of this research is to develop a practical and efficient tool for accurate tree inventory and valuation, addressing the critical need for sustainable forest management. Leveraging AI and smartphone technology, the iForester app was anticipated to deliver precise DBH and height measurements through contactless methods. These measurements are fundamental, paving the way for future functionalities such as species identification, tree grading and biomass estimation, ultimately supporting informed decision-making and promoting ecological and economic sustainability.
In 2023, iForester released version 1.0 of its AI-assisted smartphone app, capable of accurately measuring tree DBH using LiDAR and RGB data. In 2024, the focus shifted to enhancing mobile AI-based segmentation algorithms and developing a height measurement method. Researchers from Purdue University tested these advancements using diverse tree data from Indiana, combining field observations with iPhone video scans for analysis. The team assessed segmentation and height estimation accuracy, identifying influencing factors. Within researchers’ expectations, the AI significantly improved the segmentation accuracy (Figure 1) and tree height can be accurately measured using a photogrammetry-based method to overcome the limited measurement range of the iPhone LiDAR sensor (Figure 2). Using mobile Segment Anything Modeling (SAM), 547 out of 562 trees were successfully segmented, over 97% accuracy. For height measurement, experimental results demonstrated an error margin within 1 foot for the first 16 feet of the log.
The findings are crucial, as DBH and height measurement form the foundation of the project. Improved segmentation leads to more accurate DBH and height measurements, while the completion of the height measurement algorithm establishes the data processing workflow, guiding the design of the corresponding user interface and data-saving formats for practical use. The next phase of research focuses on integrating height measurement and AI-based segmentation algorithms into the iForester app. Once the height measurement module is complete, the focus will shift to tree grading algorithms for evaluating log values using DBH and height information. Concurrently, efforts are underway to develop a species classification algorithm based on tree surface vein patterns and size data.
“The iForester app provides landowners and forestry professionals with an efficient tool to streamline tree inventory and valuation, enabling precise DBH and height measurements through contactless methods,” principal investigator Song Zhang said. “This innovation replaces labor-intensive traditional practices, improving accuracy and decision making. By later integrating features such as species classification and tree grading, the app supports sustainable forest management and resource optimization. Its adoption can enhance community forestry practices, contributing to ecological and economic sustainability by helping stakeholders better estimate tree value and manage resources effectively, ultimately delivering financial benefits and fostering ecological stewardship.”
“iForester” App is available at the App store; detailed information is at the website: https://digitalforestry.org/iforester/
Figure 1: Enhanced tree trunk segmentation achieved through different prompt types (point and box) using mobile SAM.
Figure 2: A key frame of the field tree captured from the phone scan, with a nearby rod used for measuring tree height during field observations.
Goals:
The goal of this research is to develop an artificial intelligence assisted smartphone app, capable of delivering precise, contactless measurements of tree diameter at breast height (DBH) and height, By leveraging AI and integrated LiDAR and RGB data, the team anticipated significant improvements in tree trunk segmentation accuracy and predicted that photogrammetry-based methods would overcome the range limitations of iPhone LiDAR sensors for height measurement.
Methods:
The team gathered tree data from Indiana using a combination of field observations and iPhone video scans with LiDAR and RGB data. They employed mobile AI-based segmentation algorithms and utilized a photogrammetry-based method to capture and analyze tree measurements. The collected data was then used to study segmentation processes, height estimation and the factors influencing these methodologies.
Key Findings in 2024:
Segmentation: The mobile AI-based segmentation achieved 97% accuracy, with 547 out of 562 trees successfully segmented using the Mobile-SAM algorithm.
Height measurement: A photogrammetry-based method demonstrated an error margin of within one foot for the first 16 feet of the log, overcoming the iPhone LiDAR sensor’s limited range.
Future Research:
- Integrating the height measurement and AI-based segmentation algorithms into the iForester app to enhance usability and functionality.
- Developing tree grading algorithms to evaluate log values using DBH and height measurements, providing valuable insights for forest resource valuation.
- Creating a species classification algorithm by analyzing tree surface vein patterns and size data to further support forest management decisions.
Key Collaborators/Partners:
This project unites a multidisciplinary team of researchers from Purdue University, leveraging expertise in mechanical engineering, computer graphics technology, forestry and design.