Hello friends, and welcome to the fourth lecture.
And in the previous lecture I have discussed different types of a vector data, if you recall
that I mentioned the data and GIS can be divided in two major components; one or two major
types of data; one is spatial data, other one is non- spatial data, we also call them
attribute data. Then we discussed different type of vector data,point data, line or polyline
data, polygon data. Now, in this lecture we will be discussing
about what is basically raster data, and types of raster data as well. So, raster data is
having two types; one is the grid type data, another one is the image data. There are no
much differences between these two, but still the differences between these two, must be
understood very carefully, because otherwise while analysing the data, or handling such
data may give you such problems. So, understanding about the intricacies of two different types
of raster data, is very important. We will be also comparing, seeing the advantages associated
vector data as well as raster data, and this we will be doing little later part of our
lecture. So, first the raster data as you know that as a continuous data, vector data
as I mentioned is a discrete data. So, there are two different way of representing the
real world; one it is in a discrete manner using point line or polygon, may be in the
TIN model which we will discuss later, but the topic which we are going to discuss in
detail, is the raster, in a continuous fashion we discrete, we represent the real world. Now, in raster what are things which are there,
or the key things which are there, it is a uniform grid.You can imagine that if it is
a real world scenario, then a uniform grid is lead out on this.This too,the concept of
this two is based on sampling theory which is called Senan. Senan was a mathematician
and who gave this theory which is called Senan’s sampling theory. What does it mean here in
our GIS parlance, is that if you if you represent the real world in a continuous equal size
cell grid, then each cell is representing an average value of whatever present on that
part of the earth. For example, if you if you are having an agriculture
land, within a say ten meter by ten meter area, there might be variation in vegetation
or say wheat crop or some crop, but everything is average down; say for example, greenery
or may be production or anything, is average down and one single value value is assigned
to that cell, that value becomes attribute value of our raster cell, and that is why
in case of vector, we can have several attributes or theoretically n number of attributes which
each vector entity, but with raster cell or raster pixel, we can have only one single
attribute; that is one of the major difference between vector and raster. Apart from that
raster is a continuous vector is a discrete data, discontinuous data. As here you can see that data is stored here
in a two dimensional matrix, which you are having a set of rows and columns. Now number
of rows and number of columns need not to be same, as you know the matrix in mathematics,
this is also not compulsory; that means, I can have a rectangular size of my data raster
data, or I can have square shape, but apart from these two shapes, I cannot have raster
in any other shape. However when we are representing raster for an arbitrary area, and then there
are some other concept are applied which we call as no data concept, and by using no data
concept on our screen or for our visual, we see arbitrary boundary for raster. In fact,
in the system for the outside area it is stored as no data, and for inside that polygon it
is stored the real data. So, fundamentally the raster overall shape
of the raster will have to be always either rectangle or a square, because it is representing
mathematically a two dimensional matrix. Now each cell in case of grid we call as a cell,
in case of image we call as pixel, we will see little later all these details. So, each
cell will represent one attribute one value it may be in case of remote sensing data,
it may be reflection value, it may be emitted value, it may be temperature value, or in
normal case may be rainfall value or any other value, but only single value per cell or per
pixel, and this cell or pixel becomes the unit of our grid or raster. And another important
thing has to remember that the shape of this unit has to be square. So, overall shape of
a raster can be square or rectangular; however, shape of a unit has to be always a square. So, that we maintain the uniformity. Now as
a though mathematician worked on this, and rather that constraining that each cell each
unit has to have the same shape, they came with a new idea that instead of a square we
can have triangles, of varied size and shape, and that brought the concept of TIN which
we will see little later. So, that a unit of raster is always is a square; overall shape
of a raster can be square and rectangular, and each unit of raster will have only one
single attribute; not like vector data. So, raster is a continuous, and we are, generally
familiar when we are handling the image raster,; then there are image formats like tiff gpeg
and gif and different image processing software, are having their own format; like erdas is
having its own image format which is very popular one is IMG format. So, there are you
know there the data this is how it is represented and then the cells are having. Now the size of a cell can vary, it is not
always one meter one by one meter or two meter by two meter, but the shape is fixed which
is square, but the size of the cell can change larger the area, the cell is covering of a
raster. We say a smaller the spatial resolution, or poorer or courser the spatial resolution.
Whereas a cell if it representing a very small area of the ground, then we call as a higher
spatial resolution. For example, in a satellite image a one pixel might be representing ten
meter by ten meter, and we if we compare a another satellite image which might be representing,
a pixel might be representing an area of one meter by one meter; that means, the image
which is having one meter cell or one meter pixels, is having higher spatial, relatively
higher spatial resolution. Spatial resolution term is really relative,
for someone, for some applications hundred spatial resolution may be useful, but for
another person even one meter spatial resolution may not be useful. So, we have to be thinking
in this direction that is a relative term, as per our requirements we call as a higher
spatial resolution, or we call. About the spatial resolution how a different spatial
resolution will bring the changes in our image which we will see little later in this in
this presentation. Now not only the satellite images which are
handled in GIS, but there are other raster data, like a digital elevation model which
is which is representing the undulation, or topography of a terrain, or part of the earth;
that means, where are the higher grounds, lower grounds, valleys, hills, slopes and
all those things, can also be represented in a form of grid or in raster and each cell
of that grid will be representing elevation value
that is why it is called digital elevation model. Example is given here, and here that
attribute of that cell of that raster or grid going to be your elevation value. Whereas in case of image it might be having
reflection value, or emitted value, or temperature value, or any other value is represented there,
but in satellite images since this is coloured image; therefore, it is constructed using
three primary colours, and therefore, you are seeing in colour, but still all are raster,
and different type of raster which is a your image raster. So, as I have already mentioned earlier that
there are both, you know both are image and grid both are raster model, however there
are differences in case of image, the unit we refer as pixel, pixel in abbrevation, which
is pixel element, and a the cell in case of grid the unit we call as cell. This makes
a while you know calling differently and also gives gives us advantage of understanding,
which is required for the next property, or next characteristics, which is the value of
the image pixel or value of the cell. In case of image the value of a pixel can only be
positive integer value. We cannot have no other value and; that is one another big difference
between an image and a grid. Whereas, in case of grid the cell value; that means, the unit
value, can be either negative or positive integer or real numbers. So, variety of values
can go in case of grid data, but in case of image data only positive integer values. So, these two major differences are there,
when we say pixel; I am that means one is talking about image and once a person is talking
about an image that the unit value of that image; that means, the pixel value has to
be a positive integer value. And when I say cell value; that means, I am talking about
grid, and the cell value can be either positive negative integers or real values. So, the
grid is having more the having more capabilities or bringing or handling all types of all types
of integer and real numbers; whereas, the image can only handle the positive integer
number. So, this difference has to be remembered, because if we don’t a take care about this,
then because the format for grid are different, then format for image. Image formats we are familiar because now
a days we use these digital cameras, and which are also inbuilt in mobiles most of the time
we are storing images in JPEG, but we can change the format, we can store the image,
and all are these image formats and; that means, we are storing pixel values as positive
integer values. If you get sometime you can open some images, zoom it and read the values,
you would always find a positive integer value. Now, the next thing is which I test little
bit about spatial resolution, which is very very important to understand. Because a this is related, though it’s a
related term, it plays a very important role in our day to day life as well as in GIS system.
And as we go by the definition is a major of how closely pixels can be resolved in an
image; that is called spatial resolution, and it depends on the properties of the system
creating the image. Now if a what does it mean that if there are two neighbouring pixels
are there, and if they are having very you know a contrasting properties, they might
appear in a particular scale, they might appear separately. So, if you take a digital image,
may be taken by your mobile camera putted in computer zoom it to a very high level,
after sometimes you might realize where the two contrasting things are there, you might
realise, two pixel or neighbouring pixels are showing different colours or states. So,
when you reduce the scale, or zoom out you know then what happens that these starts resolving,
and then individual boundaries of the pixels you will not see. Now, in a in our domain that is in GIS or
in remote sensing, this is what we say the clarity of an image, which is decided by the
spatial resolution; that means, a an image which is having very high spatial resolution,
should be looking much pleasing to eyes; that means, might be that it is having higher spatial
resolution; that means, the number of rows in order to cover the same area, number of
rows and columns are going to be more, and; that means, the ground area which that high
resolution image is representing; that means, one pixel of that image is representing is
very very small, comparing to image which is having relatively low resolution. So, sometimes
we like in digital mobile cameras we say; ten mega pixel, twelve mega pixel, eighteen
mega pixel; that means, we are talking about the dimension of the image. So, larger the
mega pixel, the clarity of the image would be higher, higher the spatial resolution. The example is given here, that here the spatial
resolution is 110 by 110 kilometre and you are unable to resolve adjacent pixels; and
therefore, you are seeing the boundaries of pixels it is a relatively very course image,
but when you make this representation is a using a finer grid, and the having a resolution
of thirty kilometre by thirty kilometre cells, then at least a some clarity has been introduced
in this middle stage, but if a if a you improve further, and then a individual boundaries
of two neighbouring pixels are resolved then you will start seeing a much more realistic
image; and that is relatively having much higher resolution image then the right most
image. Same thing here that, this is one by one;
that means, there is nothing no differentiation is there one pixel only. Whereas, here two
by two; that means, there are total four pixels, all are having different colours, but we cannot
make out what it is written there, or what is what object is within these four pixels,
but once we start moving and you know segmenting this area into finer size pixels an image
start appearing. Here in 10 by 10 it appears may be r may be a alphabet, but here it starts
appearing as a r much clearly, but by 50 by 50 and 100 and 100 becomes much sharper. So,
the right most here, is having very high resolution, relatively as compared to the first two three
four or five. So, this is this is the major difference which resolution, especially spatial
resolution causes in our representation of raster data, or that might be a grid representing
a digital elevation model, or might be a satellite image, or photograph taken by your digital
camera. Now as I have already mentioned that vector is a discrete raster is a continuous
model. So, now, we will try to compare and see what
advantages are associated with vector data, and what other advantages are associated with
raster data, that these data sets are fundamentally completely different; vector and raster. And if a real world has to be represented
in these two formats then you can see that if real world is represented, which is a very
simple world of a real world that you are having a stream, you are having a say agricultural
land in the background and forest and a house. If I have to represent that one in the vector,
then the house is going an area here, depending on the scale in a smaller scale that house
might appear as a point data, but here as a polygon, and same with the forest is a polygon.
Whereas, a stream though in this state image, it is looking that the stream is occupying
an area, but here I am representing as a line feature; that means, only length no area.
And if I have to represent the same real world into raster, then house will have this much
size; one pixel is representing house, few pixels are representing forest, and then few
pixels or cells are representing steam or the river here. Now you will realize that I can increase number
of nodes in case of vector, and can make my line very smooth, but in order to make my
line smooth, or a steam representation much more smooth, then I need to increase the spatial
resolution; that means, the representation through my raster has to be with very small
cells, and that means, I am going to occupy more space on my computer, if and more handling
or processing time, if I go for higher and higher spatial resolution. So, it depends
on the your project requirements, Nowadays all types of all types means, all variety
of spatial resolution data are available.For example, if I take digital elevation model,
and I take the example of freely available digital elevation models, then starting from
one kilometre spatial resolution digital elevation model to up to 30 meter spatial resolution
digital elevation models are available. Now but what is the purpose if you are going for
a small are area and you are going in much more detailed analysis, then it is always
better to go for higher spatial resolution digital elevation model. But if you are covering a very large area
a large region or a continent, then it is not worth going for a high spatial resolution,
it is better to go for relatively courser of spatial resolution may be ninety meter,
may be one kilometre resolution. So, it depends on the requirements of your project and what
kind of products you are going to generate after all analysis in GIS. So, variety of spatial resolution, digital
elevation model, same with the satellite images nowadays satellite images of 30 centimetre
resolutions are available, and satellite images of one kilometre resolutions are also available.
So, again if you are looking for a if you are doing analysis on continental scale, then
you need not to go for thirty centimetres, because the data handling is going to be very
use, and you may not be able to handle data on a one single machine, using standard stand
alone GIS software, having a very high spatial resolutions data for a continental areas.
So, it is always find out what is the best and then bring in your GIS software and do
the analysis. All types of spatial resolution data nowadays are available. Now vector data models as we know that it
keeps the data either point line or polygon, and which is a vector data for certain types
of features is good, which are having discrete, but a certain types of things which are varying
continuously, like for an example given a soil, may not be good to represent through
the vector data, because soil varies in nature a very gradually, not suddenly, but if somebody
is using for lithological maps then we know, that lithology changes suddenly sometimes,
and therefore, may be vector may be useful in that way. So, raster whereas, raster evolved is a continuous
data, continuous features, both models are having advantages and disadvantages which
we will see very soon, and the modern GIS are able to handle both types of data model,
or rather three types of data model vector raster and TIN. Another important thing is
that a GIS are nowadays capable of transforming, or converting your raster to vector and vector
to raster; however, with some limitations. So, little later we will be also seeing, how
we can convert from raster to vector, and vector to raster. So, if we start looking
the advantages first with the vector raster data. This is the most common and very simple to
understand raster data easy to perform, because mathematically it’s just simply a two dimensional
matrix, and therefore, whatever the analysis tools which have been developed in matrix
domain of mathematics, all can be implemented easily in GIS. Secondly, all satellite images are also raster,
and there are many functionalities, algorithms and every capabilities have been developed,
in digital image processing software’s, and therefore, the same thing have also been
implemented into GIS. So, raster, because concept wise it is simple, and therefore,
lot of tools are available to analyse the raster data. I have already said about the
satellite data, because the satellite data becomes one of the very important data input
into your GIS, because it is easily available, it is a continuous data and with almost you
know every day you can have some satellite data, or another. Whereas, the advantages
is vector, the biggest advantage with vector is accurate, because even for a house which
might be in a small scale map, might be represented as a point, but in raster it will have an
cell; that means, area. So, even a point will have an area whereas, as per the definition
in vector point is a zero dimensional entity. So, raster is not that accurate as vector,
positional accuracy is very high in case of vector data, and it is compact because it
is discrete. There is no redundancy in the data. In case of raster you are having lot
of redundancy, and then therefore, there are data compression tools are available, to remove
that redundancy, or reduce the size of for storage of raster data, but for vector data
there is no redundancy. Once the topology has been built then redundancy is completely
gone, and no compression other things are required. So, vector data is very compact,
and large area data in vector format can be stored on a small computer space, and can
we have has another advantage that in raster you can have only one attribute, but in case
of vector you can have multiple. So, that is another advantage. As we know that raster which we have already
covered this part, the raster divides into grid cells vector uses point line polygon,
units in case of vectors are point line polygons, in case of raster, either cell or pixel. And
this processing can be done with grid and convert it from vector to raster, and all
modern GIS software’s supports all three types of data handling. Raster has we go finer and finer in the grid
cell or pixel size, we go for higher and higher resolution, and later on also, the size can
be changed, but resolution spatial resolution cannot be real, in real sense cannot be improved,
but on computer point of view you can create more number of rows and columns and therefore,
a smaller size cells are there. Now, let me take an example of a soil map. As I have said
that soil map, soil is a continuous varying feature, but sometime we have to represent
as a polygon map. So, example here is different types of soil map is represented in a polygon
map, but if the same soil map, if I have to represent in a raster, then I have to represent
something like that. So, both vector and raster representations are there. And what you will see the boundary line of
two soil units in case of raster, you are seeing a stair steps case, or also called
edge effect of the boundaries; that means, the boundaries are not sharp, and as mentioned
earlier in case of vector, the positional accuracy is very high, relatively very high
as compared to raster which you can also realize. So, on the edges of two different units, the
boundary will have this edge effect or stair steps case, something like stairs which are
there, but if you improve the spatial resolution that if you go for final spatial resolution,
then it might not be that much obvious, but anyway at every state of raster, at the boundary
then it is going arbitrary then you will have this edge effect. So, that is one of the disadvantages
with raster. So, finally, what I am going to do. I am going
to run through all these things with the raster versus vector, using different characteristics
one by one, and see that what are the advantages with vector and raster. First we characteristic
which we take is the data structure. By now you must have realized that the vector, because
of topology and different types of vectors, relatively vector data is little complex than
raster, because raster it is easy to understand, it is easy to handle in computers. So, raster
is very simple data structure compared to vector data structure, and therefore, anything
which is simple is much easier to learn. So, the ease of learning is also accordingly,
that a vector is having difficult. Software’s are complex and expensive, whereas,
raster one are much simpler. Positional precision as discussed already that a vector provides
very high precisions, and when your data representation is very accurate, precision wise, accuracy
wise then you will have a better outputs. In case of raster the output also becomes
little difficult. So, precision increase with increase processing time, and data storage,
also precision here will also increase with increase of a spatial resolution. Now, attribute
precision it is very good in case of vector, and you can have n number of attributes associated
with each vector entity point line polygon, but in case of raster it is limited only to
one entity, or one attribute value. Compressiveness of analysis capability, it
is good for a special query with some limitations, when it comes for intersections, especially
when you are overlaying two polygons together. We will say some examples of such limitations.
How these have been resolved in GIS that we will also see, but still it is limited in
GIS. Whereas, in case of raster; no issue, no problem, because these raster will stack
one of another one thing, and if they are belongs to the same coordinate system and
same area, then even the spatial resolution is different, does not matter they will stack,
and you will not find any difficulty while analysing such raster data layers. The another characteristics which is overlaying
ability with vector is good, but after sometime you start seeing splinters; that means, lot
of polygon will come after overlaying four five polygon layers, and interpretation
or uses of such maps becomes very difficult, but in case of raster that problem will never
arise. Now storage requirements as we have already realized that vector require little
less storage. it is a discrete data it does not require much storage, and therefore, though
it is complex, but it is easy it requires less space. Whereas, raster is simple, but
it requires more space, because of redundancy in the data, and when we will discuss the
raster data compression techniques then we will go in detail, how what basically redundancy
affects the raster data, ability to work with the image data. It is a not as good because
image data is raster and, when you are working with the image data and vector data you will
have problems. whereas, if you are having a image data image data is also a raster,
your grid data is also raster, no problem whatsever. Conversions to other map projections map projection
we will discuss in another lecture. So, when we go for different map projections, vector
is very good, and most of the GIS modern GIS software’s include change of projection
from one projection to another. And if you are having data in vector format it is very
easy to convert from one format to another, one projection to another, but with raster
data it is rather difficult, and it becomes, note the accuracy part will also come in between. Now ability to work with the network data
structure, because vector data why I say, because line data I gave the example that
most of this networks, like power grid or sewage line or telephone lines, ofc lines
all are networks, and they are stored in computer or in GIS software’s as vector data, and
therefore, if you are working in that domain then it is very good with the vector data,
but it is with the raster; obviously, it is very very poor. Cost expensive vector is expensive,
because each and every node has to be digitized; either wise some semiautomatic method, because
yet there is no fully automatic method to convert vector data from analogue to digital. So, maximum the best possible methods are
available, semiautomatic methods with the human interventions, lot of human intervention,
and therefore, and the vector data is expensive. Whereas,raster data is much easier, and that
is why lot of free digital elevation models which are in grid format, which is raster
free satellite images are available on the net now a days. So, it is not that expensive,
because in one go a satellite can cover a large area, but it is not possible with vector
data, and also the output map quality. Here now vector is having advantage
very good and looks like a map; whereas, your raster are poor and does not look like a map. Another problem with a raster data, vector,
because a laymen or normal people, those who do not know much about GIS or a such a fields,
they are familiar with the vector data maps. They have being seeing since very long, maps
which are having vector data, but people are not very good on seeing or interpreting satellite
images or raster data. So, that is that is why many times though we may be doing lot
of analysis using raster data, but in order to make it is much simpler we might be converting
finally, into vector. So, that our map quality becomes much much good or better, and looks
likes conventional map; otherwise a raster data, those who understand raster data for
them there is no issue. Now, I want to cover this about the overlay ability, that I mentioned
here that vector is having vector is having limited overlay. So, what does it mean here? That if I take
example of two maps both maps are having three units each, so here there are three units
a b c, and in land use map there are three units one two three.It’s a very simplest
example in real GIS operations this kind of simple operations are not perform many complex
operations, and may be that you might be having node three units, might be 30 units in one
map or 60 units in one map, but anyway in this example, which is very simple example,
attribute one set of attribute which is for a b c, the depth of soil is given here. Whereas,
in case of land use which code are or id’s are for different polygons;
are one two three, and what kind of land use in different unit of polygons are used for
as agriculture and pasture. When we overlay this one on another, using like a transparent
sheet, using your set theory concept or Booleans logic, then this kind of intersection map
will have seven units now in this particular example. So, we started with three units, and we ended
up with the seven units, seven polygons. Some have become very small, and imagine that you
are having thirty units in one map and say thirty or forty units, or polygons in another
map and such overly operations then you perform then you will end up with hundreds of such
polygon, and therefore, the interpretation of such maps becomes difficult. In earlier
versions of GIS even overlay of two vector polygons was not very easy, but nowadays it
is not that difficult. So, one more field has been added; that means, one more attribute
has been added, and initially each has single attribute now it is one more,and this id have
also changed. So, you are having soil and land use. So, a 2 a2 is having part of a and
soil map and the second unit two of the land use map, and likewise you can have overlay
even in vector data, to certain limits. But if you overlay many vector polygon vector
maps and then you will end up with very very small small polygons, splinters, and the interpretation
of such maps or utility of such maps becomes very difficult. Now I also mention that it
is, now a days it is possible to convert from vector to raster and vice versa; however,
with some limitations and whenever you convert from one format to another, one model to another
you will introduce some errors, but only when it is most essential or compulsory, then convert
from one format to another; otherwise originally if it is available in a raster format, keep
it as a raster. If it is available in vector format keep it as a vector, because once you
convert from one format or one model to another, you are bound to introduce errors, errors
are inevitable, and as mentioned in previous lectures that error propagates in GIS. So,
our aim should be to keep the error at their minimum level anyway. So, if I want to convert from vector to raster
what I need to determine, I need have the row number and column number; and that is
what here is, L and E ,L is the line number E is the element; that means, the column number,
and these are the two functions F1 and F2 which are polynomial functions. So, again
mathematical concepts are coming here again; polynomial equations are used not only in
the conversions, later on extensively polynomial equations are used in geo referencing as well.
So, when we will discuss we will go through different orders of polynomial equations,
there also. And. So, here the input are x and y from your vector data which are your
coordinates, and using polynomial functions you will get line number and row number. Similarly
in case of raster to vector conversion which is just reverse, that we need to determine
these two coordinates appear x and y. So, again there are polynomial equations,
and l and e; that means, line number and row number will go as input, and this is what
it is, the line is the scan line or line number e is the element or pixel, x is the horizontal
coordinate in the map projection, y is the bi-coordinate or vertical coordinate, and
F1 to F4 are polynomial functions. So, this is how the conversion from one model to another
model, is possible in GIS, but as I have mentioned, any conversion is not fully transparent; that
means, it is it is going to bring some errors, because you know that now you know that vector
and raster data models are two different contrasting data models, and therefore, you are bound
to have problems when you convert from one to another format,and this brings to the end
of this my presentation. Thank you very much.
