Skip to content

In this repository, we will submit programs to establish specifications for the development of leaf-clip optical sensor

License

Notifications You must be signed in to change notification settings

JMPerrot/Specifications-Development

Repository files navigation

Specifications-Development-

In this repository, we will submit programs to establish specifications for the development of leaf-clip optical sensore dicated to the monitoring of leaf biochemical content. The PROSPECT Model is used with ANGERS database to simulate different configurations to find an optimal set of optical parameters.

Reference : Initial situation and goal

Initial situation

spectral range :

-[400-900] nm for chlorophyll (CHL) and carotenoid (CAR)

-[1300-2400] nm for equivalent water thickness (EWT) and leaf mass per area (LMA)

In this case we have the situations from which we will start. We will seek to improve the estimates before degrading the spectral resolution

Approximative goal

The approximate objective is defined by the optimized estimation of the PROSPECT inversion. This inversion uses the following spectral ranges for the estimation of the different constituents :

[700-720] nm for chlorophyll (CHL)

[520-560] nm carotenoid (CAR)

[1700-2400] nm for equivalent water thickness (EWT) and leaf mass per area (LMA)

We can observe that we have an improvement in the normalized root mean square error between these two situations. The objective of this repository is to share codes retracing the steps allowing, from the initial situation, to tend towards the optimized situation by using specific wavelengths and not extended spectral ranges

R&T

First step : Sampling stepping

the first step is to downsample the spectral ranges by gradually increasing the step. The NRMSE will be calculated to quantify the quality of the estimates.

Second step : Sampling translation

The second step will be to drag the sampling comb over the spectral range using the step minimizing the NRMSE in the first step. The goal is to see if there is more relevant information between two wavelengths than that obtained in the first step.

Third step : features selection

The third step consists in carrying out a selection among the wavelengths constituting the sampling comb in the second step. We will therefore start from all the wavelengths (N wavelengths), delete one, calculate the NRMSE, put it back and delete another. We will keep the batch of (N-1) wavelengths and start again until there is only one wavelength left

Fourth step : gaussian filter application

In the fourth step, we will apply a Gaussian to the batches of wavelengths identified in step 3. This will make it possible to simulate a source of more or less good resolution. In addition, this will allow us to conclude as to the resolution necessary to have an estimate of acceptable leaf constitution.

R only with prior N estimation

N is the number of leaf structure in the plate leaf model

First step : Sampling stepping

Second step : Sampling translation

Third step : features selection

Fourth step : gaussian filter application

T only with prior N estimation

First step : Sampling stepping

Second step : Sampling translation

Third step : features selection

Fourth step : gaussian filter application

About

In this repository, we will submit programs to establish specifications for the development of leaf-clip optical sensor

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published