Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Accurate and Efficient Urban Street Tree Inventory with Deep Learning on Mobile Phone Imagery

Asim Khan, Umair Nawaz, Anwaar Ulhaq, Iqbal Gondal, Sajid Javed

TL;DR

This work addresses the need for accurate, scalable urban street-tree inventories by a mobile-centric pipeline that segments trunks with SegFormer on smartphone imagery and estimates DBH through a pinhole-model-based calculation using two consecutive views. The method removes reliance on specialized equipment, leveraging cloud-augmented inference and a two-image capture protocol to compute DBH with low error ($RMSE \approx 2.1$ cm, $MAE \approx 1.6$ cm) across three species common in Dubai, while achieving an IoU of $87.4\%$ and pixel accuracy of $93.8\%$ for trunk segmentation. A 400-image dataset from Dubai streets (Phoenix dactylifera, Vachellia nilotica, Ziziphus mauritiana) was used to train and evaluate the approach, with a ResNet-backed SegFormer model trained for 150 epochs and evaluated via standard metrics. The study demonstrates practical potential for urban forest management, enabling affordable, accurate tree inventories and informing ecological and climate-resilience planning; future work includes automated distance estimation and broader deployment.

Abstract

Deforestation, a major contributor to climate change, poses detrimental consequences such as agricultural sector disruption, global warming, flash floods, and landslides. Conventional approaches to urban street tree inventory suffer from inaccuracies and necessitate specialised equipment. To overcome these challenges, this paper proposes an innovative method that leverages deep learning techniques and mobile phone imaging for urban street tree inventory. Our approach utilises a pair of images captured by smartphone cameras to accurately segment tree trunks and compute the diameter at breast height (DBH). Compared to traditional methods, our approach exhibits several advantages, including superior accuracy, reduced dependency on specialised equipment, and applicability in hard-to-reach areas. We evaluated our method on a comprehensive dataset of 400 trees and achieved a DBH estimation accuracy with an error rate of less than 2.5%. Our method holds significant potential for substantially improving forest management practices. By enhancing the accuracy and efficiency of tree inventory, our model empowers urban management to mitigate the adverse effects of deforestation and climate change.

Accurate and Efficient Urban Street Tree Inventory with Deep Learning on Mobile Phone Imagery

TL;DR

This work addresses the need for accurate, scalable urban street-tree inventories by a mobile-centric pipeline that segments trunks with SegFormer on smartphone imagery and estimates DBH through a pinhole-model-based calculation using two consecutive views. The method removes reliance on specialized equipment, leveraging cloud-augmented inference and a two-image capture protocol to compute DBH with low error ( cm, cm) across three species common in Dubai, while achieving an IoU of and pixel accuracy of for trunk segmentation. A 400-image dataset from Dubai streets (Phoenix dactylifera, Vachellia nilotica, Ziziphus mauritiana) was used to train and evaluate the approach, with a ResNet-backed SegFormer model trained for 150 epochs and evaluated via standard metrics. The study demonstrates practical potential for urban forest management, enabling affordable, accurate tree inventories and informing ecological and climate-resilience planning; future work includes automated distance estimation and broader deployment.

Abstract

Deforestation, a major contributor to climate change, poses detrimental consequences such as agricultural sector disruption, global warming, flash floods, and landslides. Conventional approaches to urban street tree inventory suffer from inaccuracies and necessitate specialised equipment. To overcome these challenges, this paper proposes an innovative method that leverages deep learning techniques and mobile phone imaging for urban street tree inventory. Our approach utilises a pair of images captured by smartphone cameras to accurately segment tree trunks and compute the diameter at breast height (DBH). Compared to traditional methods, our approach exhibits several advantages, including superior accuracy, reduced dependency on specialised equipment, and applicability in hard-to-reach areas. We evaluated our method on a comprehensive dataset of 400 trees and achieved a DBH estimation accuracy with an error rate of less than 2.5%. Our method holds significant potential for substantially improving forest management practices. By enhancing the accuracy and efficiency of tree inventory, our model empowers urban management to mitigate the adverse effects of deforestation and climate change.
Paper Structure (9 sections, 2 equations, 7 figures, 5 tables)

This paper contains 9 sections, 2 equations, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Proposed Method
  4. Experimental Setup
  5. Dataset preparation
  6. Model Training
  7. Application Inference
  8. Results Analysis
  9. Conclusion

Figures (7)

  • Figure 1: Distance Calculation through Multi-Image Acquisition Approach and Visual Tree Projection Observation on Camera Sensor
  • Figure 2: A Workflow Diagram for Tree Trunk Diameter Estimation
  • Figure 3: The dataset includes sample images captured at different distances. The images in the first row, namely Phoenix dactylifera, Ziziphus mauritiana, and Vachellia nilotica, were taken at a distance of 20 feet. On the other hand, the images in the second row, also consisting of Phoenix dactylifera, Ziziphus mauritiana, and Vachellia nilotica, were captured at a distance of 15 feet.
  • Figure 4: A Block Diagram for Tree Trunk Segmentation and Diameter Estimation
  • Figure 5: Mobile Interface for DBH Estimation and Tree Trunk Segmentation:a) Image Pair Acquisition, b) Tree Trunk Segmentation with SegFormer NEURIPS2021_64f1f27b, c) DBH Estimation and Display of Calculated Value
  • ...and 2 more figures