Month: May 2008

Hyperspectral For All ; Soil Mapping

My research in hyperspectral technology for mineral mapping surely almost meets the goal. I try to modify some of the method for spectral classifications both in ENVI and my own method based on similarity with the spectral measurement in ground truth samples.

I like hyperspectral finally since it is state-of-the-art of remote sensing recently. I like the hundred bands for smaller range of wavelength. Yet, every subject in the earth or wherever in the universe has their own specific characteristic. The sensitivity of object might record and translated into historgram which shows into graphic displaying relativity between reflectance feature and wavelength.

The design of spectrometer must be the same with the designing of hyperspectral instrument which is carry out mostly in airborne. Currently, NASA has evaluated the launching for spaceborne hyperspectral with hundreds channel of wavelength called Hyperion. In addition, German has finished for the final assessment to the hyperspectral satellite. To be honest, those are good news. Spaceborne is able to capture wider areas and yet, reduce the cost of operation and time. I heard from Rob Hewson, remote sensing geologist work for CSIRO Exploration and Mining that the Hyperspectral satellite will launch approximately in 2012.

The other good news that the research that I am working now has been published and written by Tom Cudahy, the geoscientist and the cheap of Spectral Sensing Technology Research Group in CSIRO Exploration and Mining. Even, it is not really my work since it is the kind of joint collaboration from the fieldwork, laboratory analysist, modeling until finishing. I am quite happy that Tom enclosed my name and Curtin University for my institution as the co-author. In fact, there are nine people involve in this paper. The paper is about kaolinite mapping using Hyperspectral for guiding soil mapping. It is quite surprising and interesting since the soil mapping method is taken from conventional method which based on the field sampling collection and geochemical analysis of soil content. Tom said in the paper that kaolinite abundance that map from Hyperspectral image can help to classify the soil properties in the large areas. The access to the targeted places sometimes might be a problem. However, the kaolinite occurrence helps to select the priority areas for further investigation and sampling.

I remember when I try to map clay minerals abundance in TickHill region as my test sites for my project. The spectral feature for clay minerals shows the close similarity and concentrate in 2200nm absorption. May be only kaolinite display the different pattern compare to the other clay minerals such as illite,smectite (montmorillonite), or mica(muscovite). Kaolinite has minor absorption at 2160nm and major absorption at 2200nm. Illite and smectite appear the same feature nearly at 2200nm. So, in order to map kaolinite, I have to masking out for the clay absorption at 2200nm. Amazingly, the kaolinite kristalinity can be produced by masking out the major absorption. Then, we can say that the map that I produce can be recommended as a kaolinite for soil classification.

Latest discussion with Tom is comparing the “black box” spectral classification method which is packaged in wizard namely Spectral Hourglass Wizard. It is the automatically wizard to derive spectral classification in ENVI. At the end, it will produce three main spectral classification : SAM (Spectral Angle Mapper), MTMF (Mixture Mixing) and spectral unmixing. The spectral abundance then compare with band threshold spectral image for clay minerals. Even for the high abundance I confirm the results. Well, for less abundance, it is pretty much hard to differentiate. Yet, the band threshold image is able to show the gradation of mineral from most likely, less likely and low abundance.

I think that mineral mapping can be a guide for the next exploration project beside geophysic method such as geomagnetic, and such kind of reconnaissance study for exploration.

Anyway, for my discussion with Tom, he suggests me to compare this method and choose the most effective method that is mostly meet the target. Then, I asked him to generate chlorite and carbonates mineral with band threshold method.

May be next Friday I am going to meet him and discuss about that. Now, I have to understand the theory behind the mineral classification method which mostly based on statistical approaches.

Following the instruction in the software is easy. The friendly gadget allows me to explore some of the Spectral analysist from the Spectral tool in ENVI toolbar. Again, I need to understand theoretically. Otherwise, it was like black box without understanding the meaning.

During discussion Tom asked me what I am going to know with the Hyperspectral analysist. I said I want create the specific formula and also develop my own formula for mapping the minerals based on its spectral characteristics. Well, he said that if we know exactly what the minerals are and their characteristic on the spectra and it also confirm with the laboratory analytical mineral. It easily creates the formula. It also is proven that some of the method develop in ENVI does not completely perfect for every different minerals. Thus, it is proven on my project.

Well, I think I finished writing for today telling about my project. Azim has coming and we plan for take dinner together.

Perth, April 9th 2008


Hyperspectral Method

Since last day, I felt sorry and nervous. You know why ? since I havent finished all my hyperspectral method for my research area. I meet my supervisor and he said ok. It means that I dont have to finish all twelve lines of hyMap script. We do here a scientific approach, it is not the consultant profession. I though what he said is right.

If I calculate from the twelve lines. I will make five mineral maps (iron oxide, kaolinite, illite, chlorite, carbonates and montmorillonite) then, each mineral has five methods to apply. So, five minerals multiply by five methods and multiply by twelve lines. What !!!, I will spend all the day see the reflectance patterns and adjust with the spectra library..Ooh..No way..?!! I’ve thounght it several times. that I have to create 5x5x12 = 300 files and each files contains of 2 Gb at least..Such a big file hahaa..

Finally, I just laugh at myself. I am  not the consultant and I am not paid by anybody. That what the joke when I meet my supervisor at his office. However, I have to finish 300 files if I have enough time. But, now, just take priority. Band Threshold, the art of creating treshold for mineral mapping seems like arguing something based on the experience of interpreter..Thats quite nice I reckon.

So, I made lists of questions for Tom, my second supervisor. Hopefully, he can motivate me to do more on Hyperspectral methods : SAM, MTMF, Spectral Unmixing and BandTreshold ( it;s my favourite word, simply but wortly..thanks Tom for help me with this).

Morning has broken. let see what’s gonna happen. I dont want to complain anymore. I love my research.