Antonio J. de Castro

Spectral high resolution feature selection for retrieval of combustion temperature profiles

The use of high spectral resolution measurements to obtain a retrieval of certain physical properties related with the radiative transfer of energy leads a priori to a better accuracy. But this improvement in accuracy is not easy to achieve due to the great amount of data which makes difficult any treatment over it and it’s redundancies. To solve this problem, a pick selection based on principal component analysis has been adopted in order to make the mandatory feature selection over the different channels. In this paper, the capability to retrieve the temperature profile in a combustion environment using neural networks jointly with this spectral high resolution feature selection method is studied.

Bispectral infrared forest fire detection and analysis using classification techniques

Infrared cameras are well established as a useful tool for fire detection, but their use for quantitative forest fire measurements faces difficulties, due to the complex spatial and spectral structure of fires. In this work it is shown that some of these difficulties can be overcome by applying classification techniques, a standard tool for the analysis of satellite multispectral images, to bi-spectral images of fires. Images were acquired by two cameras that operate in the medium infrared (MIR) and thermal infrared (TIR) bands. They provide simultaneous and co-registered images, calibrated in brightness temperatures. The MIR-TIR scatterplot of these images can be used to classify the scene into different fire regions (background, ashes, and several ember and flame regions). It is shown that classification makes possible to obtain quantitative measurements of physical fire parameters like rate of spread, embers temperature, and radiated power in the MIR and TIR bands. An estimation of total radiated power and heat release per unit area is also made and compared with values derived from heat of combustion and fuel consumption.

Forest fire studies by medium infrared and thermal infrared thermography

The new generation of dedicated satellites for remote sensing of forest fires now in advanced development demands validation measurements from airborne platforms. A digital image acquisition system in the medium infrared (MIR) and thermal infrared (TIR) bands has been set up specifically for imaging of fires. The system provides simultaneous, co- registered and radiometrically calibrated MIR and TIR images. This makes it possible to use image processing techniques based on pixel-by-pixel comparison of MIR and TIR brightness temperatures. Analysis of a laboratory flame and a wood bonfire shows that the MIR-TIR scatterplot can be used to classify the scene into different fire regions (cold background, hot nonflaming soil, hot flaming soil and pure flame). This technique has been applied also to observations of forest fires realized from a helicopter at distance of more than 1 Km, revealing that several fire regions can be demarcated, including a fire front in which flame emission makes a large contribution.

Fabry-Perot resonators for high-resolution infrared gas sensors

A new Fabry-Perot filter based on a silicon wafer spacer is proposed in this work. IR gas sensors based on these filters would combine excellent selectivity and signal-to-noise ratio with an overall scheme similar to that of the simplest non-dispersive IR sensors. The filter can be fitted to the fine structure of different gases, in particular those diatomic with unlike atoms as carbon monoxide among others. The spacer of the filter is a silicon wafer of defined thickness. Fine tuning to gas absorption peaks can be reached by coupling it to other multilayers. In this work the properties of silicon wafers as spacers of the proposed filters are studied. From the study it derives that the most determining factor of the filter properties is the loss of coherence caused by surface roughness. However, as we demonstrate in this work, surface roughness limitations are not very severe and easily obtained by standard polishing procedures.

Remote sensing of carbon monoxide by open-path FTIR spectroscopy: comparison of different analysis techniques

Fourier Transform Infrared (FTIR) spectroscopy is a well-established technique for monitoring air pollutants by extractive methods. Remote sensing by Open-Path FTIR technique incorporates the advantages of a non-intrusive technique. EPA and VDI have recommended some guidelines for the application of this promising technique. However, it is necessary to do more research to assess the quality of these systems on the basis of European standards. The analysis of FTIR spectra are usually carried out by using methods based on classical least squares (CLS) procedures. In this work a line-by-line method (SFIT) is additionally used. SFIT is a non-linear least-squares fitting program that was designed to analyse solar absorption spectra. For this work, SFIT has been adapted and applied to Open-Path FTIR spectra. The objective of this work is to study the capability of both methods to analyse open-path measurements of carbon monoxide. From a previous work it was inferred that the selection of the analysis spectral window is a relevant parameter of SFIT analysis. Therefore, the first step has been to analyse synthetic spectra of known concentration to select the best spectral region and other parameters of analysis. Afterwards, the SFIT software has been applied to Open-Path experimental spectra. Results of the SFIT method have been compared with the results of the two methods of EVAL analysis. EVAL is a commercial software (provided with the instrument) that is based on a CLS procedure and on the absorption peak intensity. The result has been validated by comparison to a standard extractive method.

High-resolution multigas sensors based on solid state resonant cavities

A new sensor is proposed combining arrays of interference filters with resonant cavities of high interference order (Fabry-Perot resonators) using both, high capabilities of accurate thickness control in silicon substrates derived from microelectronics and thin-film optics techniques. The interest of this type of device would be the fabrication of high resolution multigas sensors, compact, with no moving parts, for accurate measurement of low concentration gases.

Remote sensing of water clouds temperature with an IR camera on board the ISS in the frame of JEM-EUSO Mission

The Extreme Universe Space Observatory (EUSO) is an astronomical telescope that will be hosted by the Japan Experiment Module (JEM) on the International Space Station (ISS). The telescope will determine Ultra High Energy Cosmic Rays properties by measuring the UV fluorescence light emitted by the particles generated in the interaction between the cosmic rays and the atmosphere. Therefore, cloud information is crucial for a proper interpretation of the data. To obtain the cloud top height an IR camera is being designed. The design is constrained by JEM-EUSO requirements which are mainly the instrument weight, power and data rate. These requirements have led to a bi-spectral camera option with 1 μm-wide bands centered at 10.8 and 12 μm. The bi-spectral design has allowed us to develop a Split Window Algorithm to correct the atmospheric effects and retrieve the cloud temperature from the brightness temperatures in the bands aforementioned. The algorithm has been checked in synthetic scenarios at pixel level. The simulations consider clouds at different levels with diverse atmospheric conditions. The results show that the algorithm is able to retrieve the temperature with accuracy much better than the required one by the JEM-EUSO mission of 3K. It has also been tested in 2D real scenarios (MODIS images). The algorithm has been applied to MODIS brightness temperatures in bands 31 and 32 which are similar to those of the IR camera. The temperatures retrieved by the algorithm are in a very good agreement with the cloud top temperatures given by MODIS.

Spot shape and size on the focal plane of specific infrared nonimaging sensors for the detection of forest fires

This paper describes some of the work performed in the course of the design and development of a new IR sensor system for early detection of forest fires. The proposed device is a non- imaging sensor that would discriminate angular position by means of a simple IR array, working in the 3 - 5 microns wavelength region, placed at the focal plane of the optical system. In order to accomplish low cost requirements, a system with a sole IR lens has been designed. In this work, a study of the spot shape, size and optical IR power on the detector has been performed. From the analysis of the influence of lens-detector distance and incidence angle, we have derived an optimum pixel size and optical configuration. The use of TE- cooled PbSe detectors is proposed, as well as a simplified cell array.

Discriminant regression analysis to find homogeneous structures

The main motivation of this paper is to propose a method to extract the structure information from the output data and find the input data manifold that best represents that output structure. A graph similarity viewpoint is used to build up a clustering algorithm that tries to find out different linear models in a regression framework. The main novelty of the algorithm is related with using the structured information of the output data, to find out several input models that best represent that structure. This novelty is base on the intuition that similar structures in the output must share a common model. Finally, the proposed method is applied to a real remote sensing retrieval problem where we want to recover the physical parameters from a spectrum of energy.

Machine Learning Approaches for the Inversion of the Radiative Transfer Equation

Estimation of the constituents of a gas (e.g. temperature, concentration) from high resolution spectroscopic measurements is a fundamental step to con- trol and improve the efficiency of combustion processes governed by the Radiative Transfer Equation (RTE). Typically such estimation is performed using thermocou- ples; however, these sensors are intrusive and must undergo the harsh furnace envi- ronment. In this paper, we follow a machine learning approach to learn the relation between the spectroscopic measurements and gas constituents such as temperature, concentration and length. This is a challenging problem due to the non-linear behav- ior of the RTE and the high dimensional data obtained from sensor measurements. We perform a comparative study of linear and neural network regression models, using canonical correlation analysis (CCA), principal component analysis (PCA), reduced rank regression (RRR), and kernel canonical correlation (KCCA) to reduce the dimensionality.