Annonce

3 septembre 2024

Hyperspectral image processing development for Determining tree species, Monitoring hydric/water stress Location of Volatile Organic Components zones

Catégorie : Doctorant

Diginove is a "Young Innovative Company" based in Aix-en-Provence, recognized for its work in space imagery. Its TeleCense project employs earth observation technologies to assess population distribution in a given area. This information aids in census efforts, regional development planning (including water and sanitation, energy, roads, telecoms, and transport), monitoring natural areas (such as forests, crops, and rivers), and anticipating the impact of natural disasters on infrastructure and populations.

A wide range of data is utilized for the service, with the primary source being satellite observation images. These include free images from the Copernicus constellation (radar, optical) and, when higher precision is required, very high-resolution images available for a fee.

Several new satellites will be launched soon to enhance observation accuracy:

-CO3D: Provides digital surface and terrain models using stereoscopic views from four satellites.

-Trishna and LSTM: Offer VNIR, SWIR, and TIR data for surface temperature, plant evapotranspiration, crop health, etc.

-ROSE-L: Features an L-band synthetic aperture radar for observing elements below the canopy, such as land movement, soil moisture, and vegetation volume.

-CHIME: Equipped with a hyperspectral imager for environmental monitoring, precise determination of tree species, and vegetation health.

-CO2M: Uses a spectrometer operating in the near and shortwave infrared to measure carbon dioxide emissions from human activities

Thesis objectives

These new generations of satellites will enable us to address a wide range of new applications. To prepare for their arrival, we need to:

1.Strengthen expertise in various fields of application, with the goal of developing future applications, products, and services that address societal and environmental challenges.

2.Enable rapid adoption by end-users and stakeholders of the new knowledge derived from these services, particularly in areas such as food security and agriculture, climate change adaptation (including the monitoring of forests, protected areas, urbanization, and resilience to natural disasters).

We have chosen to focus on the "forest fire" risk, which is particularly significant in our region. Our aim is to improve fire prevention by:

-Mapping crop types and tree species,

-Monitoring vegetation health, particularly water stress,

-Identifying areas with volatile organic compounds (VOCs), which can create explosive zones in the presence of fire.

For each of these objectives, the hyperspectral or multispectral imagers mentioned above can be utilized.

Research methodology

The three objectives are intrinsically linked. Determining tree species is a prerequisite for both monitoring water stress and locating areas where VOCs may be present.

Determining tree species

The challenge arises when species are mixed in the same area (e.g. oak and pine). What differentiation capability do we have, and at what resolution? This will require the use of multispectral and hyperspectral imagery, and the use of existing or future satellites.

Monitoring hydric/water stress

Throughout the fire season (which gets longer every year), it is necessary to monitor the water stress of vegetation in order to characterize fire risk. It is therefore necessary to build up time series and monitor changes in the spectral response of different species throughout the period under consideration.

Location of VOC zones

Plants react to water stress in a number of ways, including modifying their metabolism to save water and reducing growth. Some species can produce more VOCs when subjected to water stress. Other species naturally produce high levels of VOCs, regardless of their condition.

Global approach

To achieve these objectives, a systematic approach is necessary and the the following steps should be undertaken:

-Satellite(s) Identification: it involves identifying the satellite(s) capable of meeting the specific requirements by evaluating the sensors onboard these satellites in terms of acquisition revisit time, spatial resolution, spectral characteristics, and other relevant parameters.

-Data preparation and Creation: representative simulated/synthetic data must be created for the identified sensors to ensure the data accurately reflects the necessary acquisition parameters.

-Integration of non-spatial data: integrating non-spatial data sources such as LIDAR-HD, drone-borne hyperspectral imagers, IoT terrain data, and ONF databases is crucial, with a thorough evaluation of these sources for quality, refresh rate, and accuracy.

-Operational Scenario Simulation: an operational scenario tailored to the specific needs and areas of interest should be simulated by combining data from various sources, including proxy data from other operational missions, simulated or synthetic data generated using models, and in situ validation and campaign data.

-Development of Operational Services: This step involves building the initial operational services and applications based on the simulated data, providing a foundational framework that can be refined over time.

Validation and Adjustment: Finally, as new satellites are launched and real data becomes available, the simulated data should be validated against this real data, allowing for the adjustment and refinement of the operational service.

Candidate profil

M2 or engineering school graduated student in applied mathematics, signal and image processing, machine learning, computer science. Some knowledge in physics will be appreciated.

Contacts :

Ali AHAMED : Diginove (aahmad@diginove.com)

Mouloud ADEL : Aix-Marseille University (Professor) : (mouloud.adel@univ-amu.fr)

Localisation

CTO, Diginove sas (13090 Aix-en-Provence)

Campus universitaire de Saint-Jérôme (13013 Marseille)

Beginning of the thesis

January 2025

Duration

3 years

References

• VOCs hyperspectral imaging: A new insight into evaluate emissions and the corresponding health risk from industries