AsianScientist (May. 01, 2025) – Urban greenery does more than just look pretty. City plants can offset greenhouse gas emissions, reduce the urban heat island effect, and attenuate stormwater, reducing flood risks. Green spaces provide health benefits for residents by acting as recreational spaces for physical activity and social interaction.
As cities grow larger and urban environments more complex, maintaining the health of urban greenery is becoming challenging for city planners and governments.
One of the biggest challenges is that maintaining the health of trees in urban areas is extremely labour and resource intensive. It requires consistent monitoring from botanical experts over extended periods of time. To address this, researchers in Japan have developed Plant Doctor – an AI driven tool that can track plant health using video footage. Their results were published in Measurement.
“Machine vision techniques such as segmentation have great applications in the medical field. We wanted to extrapolate this technology to other areas, such as plant health,” said first author Marc Josep Montagut Marques, PhD student at the Department of Integrative Bioengineering at Waseda University.
Plant Doctor is a hybrid AI system, combining three different machine learning models to process the video footage and analyse images. Two machine vision models – YOLOv8 and DeepSort, enable efficient leaf detection and tracking in video footage. The researchers customized YOLOv8 to identify individual leaves and regions of interest, while DeepSORT tracks the leaves across video frames to choose the best images to process. A third algorithm, DeepLabV3Plus, then identifies damaged areas on the surface of leaves.
The team trained the models on a diverse dataset, including footage from plants commonly found in Tokyo. They manually annotated sick and healthy leaves to ensure the model was trained accurately.
When tested on footage of four plant species found in the Tokyo urban area, Plant Doctor was able to recognise damaged leaves regardless of lighting and shape variations in the leaves, though it struggled to identify heavily damaged leaves with wide variations in appearance.
The researchers believe that technology could vastly reduce the resources needed to monitor urban greenery, as cameras can be attached to drones and even city maintenance vehicles like garbage trucks. Using video footage instead of physical samples of leaves and branches also reduces stress on the plants. “We have provided a tool for botanical experts to assess plant health in one solution without the need to gather samples and damage the plants in the process,” said Marques.
The researchers plan to further improve Plant Doctor to allow it to differentiate damage types in foliage and improve its ability to track individual leaves and highly damaged leaves.
Beyond urban greenery, Plant Doctor could also be applied in agriculture to monitor crop health to identify diseases early, says the study.
—
Source: Waseda University ; Image: Freepik
This article can be found at Plant Doctor: A hybrid machine learning and image segmentation software to quantify plant damage in video footage
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.