Thomas Ngigi (PF 2825)
Curriculum Vitae
Personal details
Full name:
Thomas Gathungu Ngigi
Date of birth:
26th October 1970
Languages spoken:
English, Swahili, Japanese (intermediate)
Contact:
E-mail; tomngigi@hotmail.com
Tel.; +254-728786239, +254-738651033
Mail; Department of Geomatic Engineering and Geo-spatial Information Systems
Jomo Kenyatta University of Agriculture and Technology
P.O Box 62000-00200, Nairobi, Kenya
Educational background
2004 - 2007:
Institution; Chiba University, Japan
Degree course; PhD in Earth and Human Environmental Science. Certificate
Research topic; Mix-unmix Classifier: a proposal for solving under-determined
models in (linear) spectral unmixing
2000 - 2003:
Institution; Chiba University, Japan
Degree course; MSc in Image Informatics. Certificate
Research topic; Monitoring deforestation in Kenya
1991 - 1996:
Institution; University of Nairobi, Kenya
Degree course; BSc in Surveying and Photogrammetry. Certificate
Research topic; Comparison of Roelof’s prism and Sun’s limbs in solar-azimuth
determination
Work experience
2008 - current:
Employer; Jomo Kenyatta University of Agriculture and Technology
Position; Lecturer (and, MSc- & BSc-final-year projects coordinator)
1997 - 2000:
Employer; Ministry of Lands, Settlement and Housing
Position; Lecturer-cum-land surveyor (Kenya Institute of Surveying and Mapping)
Publications
A) Journal Papers
1. Ngigi, T.G.; Tateishi, R.; and Gachari, M. Global mean values in linear spectral unmixing:
double fallacy!, International Journal of Remote Sensing, Vol. 30, No. 5, pp. 1109 – 1125, 2009.
Abstract
2. Ngigi, T.G.; Tateishi, R.; Al-Bilbisi, H.; Gachari, M; and Waithaka, E. Applicability of the Mixunmix Classifier in percent-tree and -soil cover mapping. International Journal of Remote
Sensing, Vol. 30, No. 14, pp. 3637 – 3648, 2009. Abstract
3. Ngigi, T.G.; Tateishi, R.; Shalaby, A.; Soliman, N.; and Ghar, M. Comparison of a new
classifier, Mix-unmix Classifier, with conventional hard and soft classifiers, International
Journal of Remote Sensing, 29, No. 14, pp. 4111 – 4128, July, 2008. Abstract
4. Ngigi, T.G. and Tateishi, R. Solving under-determined models in linear spectral unmixing of
satellite images: Mix-Unmix Concept (advance report), Journal of Imaging Science and
Technology, Vol. 51, No. 4, pp. 360 – 367, July/August, 2007. Abstract
5. Ngigi, T.G. and Tateishi, R. Monitoring deforestation in Kenya, International Journal of
Environmental Studies, Vol. 61, No. 3, pp. 281 – 291, June, 2004. Abstract
6. Ouma, Yashon, Ngigi, T.G., and Tateishi, R. On the optimization and selection of wavelet
texture for feature extraction from high-resolution satellite imagery with application
towards urban-tree delineation, International Journal of Remote Sensing, Vol. 27, No. 1, pp. 73
– 104, January, 2006
B) Conference papers
1. Thomas G. Ngigi and Ryutaro Tateishi; Monitoring deforestation in Kenya, Proceedings of the
24th Asian Conference on Remote Sensing, Busan, Korea, pp.171-174, November 2003
2. Soliman, N.M.; Adel S.; Ngigi, T.; and R.Tateishi; Spectral Discrimination of Hydrothermal
2
Minerals Using Aster Data. Case Study Um Nar Area, Egypt, The 13th CEReS International
Symposium on Remote Sensing, Chiba University, Japan, pp.90-92, October 2007
3. Ouma Y.O.; Ngigi, T.G..; and Tateishi, R.; A preliminary investigation into the application of
wavelets transform as a fast-unsupervised environmental change detection strategy,
Proceedings of the Indonesia-Japan Joint Scientific Symposium, Chiba, Japan, pp. 47-52, October
2004
Current / future research topics
A) Theoretical
1.
Minimum and maximum values versus conventional average values in linear spectral unmixing.
Details
2.
Establishment of actual mixture model(s) to improve spectral unmixing results. Details
3.
Mix-unmix Classifier: spatial distribution of end-members. Details
4.
Understanding societal problems: from hypothesis to reality. Details
5.
Least squares method: unquestionable? Details
B) Application
6.
Forest cover map of Kenya. Details
7.
Soil-mixture map of Kenya. Details
8. Mix-unmix Concept: improvement in interpretation of medical imagery. Details
9. Potential geothermal points in the Kenyan Rift Valley. Details
Referees
3
Back
Back
4
Back
Back
5
Back
Back
6
Abstracts of Publications
Back
Authors:
T. G. Ngigi, R. Tateishi and M. Gachari
Title of Publication:
Global mean values in linear spectral unmixing: double fallacy!
Abstract:
Almost all conventional linear spectral unmixing techniques are based on
the principle of least squares. The global mean digital number (DN) of an
endmember is taken as the representative (i.e. contributory) DN for the endmember. This paper sets out to prove that the notion is a fallacy, and will
always lead to negative percentages, super-positive percentages and non100% sum of percentages if the unmixed pixel is not composed of, to
within some tolerance, the global mean DNs only. Three sets of spectral
end-members (two, three and four spectral end-members) are generated
from Landsat ETM+ data. Practical percentages (between 0% and 100%
and totalling 100%) of the end-members are returned by pixels in which
the local mean DNs of the spectral end-members do not differ from the
global mean DNs by, on average, 4.
Name of Journal:
International Journal of Remote Sensing, Vol. 30, No. 5, pp. 1109 – 1125.
DOI: 10.1080/01431160802235886
Year of Publication:
2009
Back
7
Back
Authors:
Ngigi, T.G.; Tateishi, R.; Al-Bilbisi, H.; Gachari, M; and Waithaka, E
Title of Publication:
Applicability of the Mix-unmix Classifier in percent-tree and -soil cover
mapping
Abstract:
The Mix–Unmix Classifier is a simple novel method developed to address
the problem of under-determination in linear spectral unmixing. This paper
tests the applicability of the Mix–Unmix Classifier in percentage mapping
of tree cover and different soil types from single bands of satellite imagery.
Various transformations were executed on African Moderate Resolution
Imaging Spectroradiometer (MODIS) data bands 1, 2, 3, 4, 6 and 7. The
equatorial
rainforest is most distinguishable under skewness. The
skewness transformation band is unmixed into two endmembers: tree
(endmember of interest) and non-tree (background). The resulting
percentage tree cover map was compared with a University of Maryland
percentage tree cover map of the continent, giving a correlation coefficient
of 0.87. Fraction images of three soil types were generated from Japanese
Earth Resources Satellite (JERS) synthetic aperture radar (SAR) L-band
data covering a section of Jordan. The soil types considered were hardpan
topsoil, Qaa topsoil, and topsoil of herbaceous layer. The correlation
coefficients of the Mix–Unmix Classifier-derived fraction images versus
reference fraction images for the three soil types were 0.89, 0.87 and 0.89,
respectively.
Name of Journal:
International Journal of Remote Sensing, Vol. 30, No. 14, pp. 3637 – 3648.
DOI: 10.1080/01431160802592526
Year of Publication:
2009
Back
8
Back
Authors:
Thomas G. Ngigi, Ryutaro Tateishi, Adel Shalaby, Nehal
Soliman and Mohamed Ghar
Title of Publication:
Comparison of a new classifier, the Mix–Unmix Classifier, with
conventional hard and soft classifiers
Abstract:
‘The number of bands must be more than the number of end-members…’ is
perhaps the most ubiquitous statement in linear spectral unmixing. The
Mix–Unmix Classifier overcomes this limitation. Further, the classifier
creates a processing environment that allows any pixel to be unmixed
without any sort of restrictions (e.g. minimum determinable fraction),
impracticalities (e.g. negative fractions), or trade-offs (e.g. either positivity
or unity sum). The classifier gives not only the most probable fractions of
end-members, but also their most probable contributory DNs. The
contributory DNs directly define the qualities, (e.g. the phenological stages)
of the end-members.
The classifier is applied as a dual classification method and
compared with popular conventional hard and soft classifiers in production
of two to eight spectral classes/end-members from Landsat 7 ETM+ data.
The classifiers considered are Spectral Angle Mapper, Binary Encoding
Classifier, and Maximum Likelihood Classifier for hard classification; and
IDRISI Kilimanjaro Probability Guided Option linear unmixing technique
for soft classification. The Mix–Unmix Classifier performs better than the
others.
Name of Journal:
International Journal of Remote Sensing, Vol. 29, No. 14, pp. 4111 – 4128.
DOI: 10.1080/01431160701772559
Year of Publication:
2008
Back
9
Back
Authors:
Ngigi, Thomas G. and Tateishi Ryutaro
Title of Publication:
Solving under-determined models in linear spectral unmixing of satellite
images : Mix-Umix concept (Advance Report)
Abstract:
This paper reports on a simple novel concept of addressing the problem of
underdetermination in linear spectral unmixing. Most conventional
unmixing techniques fix the number of end-members on the dimensionality
of the data, and none of them can derive multiple (2+) end-members from a
single band. The concept overcomes the two limitations. Further, the
concept creates a processing environment that allows any pixel to be
unmixed without any sort of restrictions (e.g., minimum determinable
fraction), impractical/ties (e.g., negative fractions), or trade-offs (e.g., either
positivity or unity sum) that may be associated with conventional unmixing
techniques. The proposed mix-unmix concept is used to generate fraction
images of four spectral classes from Landsat 7 ETM+data (aggregately
resampled to 240 m) first principal component only. The correlation
coefficients of the mix-unmix image fractions versus reference image
fractions of the four end-members are 0.88, 0.80, 0.67, and 0.78.
Name of Journal:
The Journal of imaging science and technology Vol. 51, No. 4, pp. 360 –
367 ISSN 1062-3701
Year of Publication:
2007
Back
10
Back
Authors:
Ngigi, T.G. and Tateishi, R.
Title of Publication:
Monitoring deforestation in Kenya
Abstract:
Multi-temporal data is used to determine the rate of deforestation between
the years 1976, 1987 and 2000. Three Landsat images, for each period, are
pre-processed, mosaicked and normalized difference vegetation index
(NDVI) values computed. Based on the values, totally non-forested areas
are masked out. The forested areas, both partially and wholly, show a very
high degree of correlation between all the bands (reflective), thus
necessitating an application of principal components transformation. The
first two principal components and NDVI values are used in K-means
unsupervised classification to distinguish forest from non-forest areas (that
appeared as forest at first). Comparison of the resulting thematic maps
gives an annual deforestation rate of roughly 15,000 ha. or 2% between any
two epochs.
Name of Journal:
International Journal of Environmental Studies, Vol. 61, No. 3, pp. 281 –
291. DOI: 10.1080/0020723032000170959
Year of Publication:
2004
Back
11
Back
Current / future research topics
A) Theoretical
1. Minimum and maximum values versus conventional average values in linear spectral unmixing.
The objective is to demonstrate that, in linear spectral unmixing, usage of minimum/maximum
DNs (digital numbers) for all end-members except an index end-member in which case the
average DN is adopted as the representative DN, gives far much better results than the
conventional usage of average DNs for all the end-members;
2. Establishment of actual mixture model(s) to improve spectral unmixing results. All conventional
spectral unmixing techniques assume linear mixing of reflectance of land covers. Hence, at the
unmixing stage, they apply a linear model. For linear mixture to occur, there should be, inter alia,
no multiple scattering within a given land cover or among land covers. However, this is rarely the
case. The object of this research is to establish the mixture model(s) associated with different land
covers within a given scene and subsequently use the model(s) at the unmixing stage. The module
will be incorporated into the Mix-unmix Classifier;
3. M ix-unmix Classifier: spatial distribution of end-members. No conventional image classifier
gives the spatial distribution of end-members. The research will involve pattern recognition in
pure pixels, and propagation in impure pixels. The former may borrow very much from current
general pattern recognition, whereas the latter has never ever been attempted in any studies;
4. Understanding societal problems: from hypothesis to reality. A case study will be carried out in
Africa’s biggest slum, Kibera in Nairobi. The slum is a perfect case of societal problems and
survival. Under the Mix-unmix Classifier concept, all possible causes/predisposing factors of
problems, as well as social parameters like sex, age, and academic standards (these social
parameters may themselves be part of predisposing factors), will be quantified and on their basis,
hypothetical problems (and their levels) generated. Practical problems will then be assessed by
back-propagating through the hypothetical problems branch to establish their causes/predisposing
factors, as well as the levels of influence of the causes/predisposing factors;
5. Least squares method: unquestionable? This will hypothesize that the concept, behind the method,
of minimising squared deviations in mathematical modelling leads to, at the very least, not the
best possible model. An alternative shall be provided;
Back
12
Back
B) Application
6. Forest cover map of Kenya. The MSc research covers only a fraction of the country due to high
cost of the high-resolution satellite data (Landsat) used. The Mix-unmix Classifier will be applied
on free/low-cost low-resolution satellite data to cover the entire country at comparable accuracy;
7. Soil mixture map of Kenya. Only 7% of the country is under crops. From the free/low-cost lowresolution images, mixture types and mixture proportions of soils will be determined nationally.
On the basis of the information and crop types, and auxiliary data, from the cropped areas, crop
suitability maps of the non-cropped areas will be produced for different crops. Crops already
existent in the country will be considered at this stage. In the areas where the crops would
potentially not fair well, especially in very arid regions, crop suitability maps of exotic crops
grown in desert areas (specifically Egypt because data on the country’s crops requirements is
available) would be produced with the aim of determining the crops whose requirements can be
met in Kenya. Also, forest/tree suitability maps will similarly be produced. The suitability maps
are to recommend types of forests/trees that can survive in potentially forestable areas;
8. Mix-unmix Concept: improvement in interpretation of medical imagery. Medical imagery
assessment is done visually. This is very subjective as different analysts may interpret the same
image differently. Further, for a tumour (or any other disturbance) to be detected, it has to be
apparent, i.e. must stand out from the surrounding areas. A tumour develops gradually and before
becoming apparent, it is ‘mixed’ with the affected area. The Mix-unmix Classifier would come in
handy in detecting the tumour at this very early stage;
9. Potential geothermal points in the Kenyan Rift Valley. Using affordable/free low resolution
satellite infrared data, a higher resolution geothermal map (after realising the above third research
topic; Mix-unmix Classifier: spatial distribution of end-members) will be produced grading the
pixels in the map according to their potential in generation of geothermal power.
Back
13
Back
Referees

Prof. Ryutaro Tateishi
Center for Environmental Remote Sensing
Chiba University
1-33, Yayoi, Inage
Chiba 263-8522
Japan
Tel.; +81-43-2903850

Dr. Edward Waithaka
Chairman
Department of Geomatic Engineering and Geo-spatial Information systems
Jomo Kenyatta University of Agriculture and Technology
P.O. Box 62000-00200
Nairobi
Kenya
Tel.; +254-67-52711

Mr. Peter Kumunga
Principal
Kenya Institute of Surveying and Mapping
P.O. Box 64005
Nairobi
Kenya
Tel.; +254-20-8561484 / 6
Back
14