Remote Sensing And Image Processing - Assignment Example

Marketing Name: Course: Instructor’s Name: Date: Question 1 Total marks for this question is 25. 1.1 Observe these 3 image band combinations of same region in Queensland, and write one advantage and one disadvantage of each combination. (15) Advantages Disadvantages A It reduces the amount of training required to make use of the imagery since the colours are familiar and they relate to physical components of the scene. 8-bit colour (256 colours) are used to represent a scene that draws (potentially) from a 24-bit (256^3 = almost 17 million) palette of possible colours. B During visual analysis, is difficult to analyse images in 3-dimension. In this band combination, it is easier to analyse and interpret the image since colours have specific meaning. It has limited area of interpretation since a lot of information will not be specified on the image. The distinct types of colours used make it hard to include other earth features when analysing. C They can be used in the fire management applications. This includes conducting analysis for post-fire of burned and non burned forested areas. It is only limited to desert like conditions and opposed to other geographical features. 1.2 Name five main land use types you may find in this image? (10) 1) Farming activities. The presence of a dark green colour is a very popular band combination that tells the presences of vegetation. They tell the presence of wet earth from dry earth. 2) Urban or settlements; densely populated urban areas are shown in light blue. The fact that the area on the image is a densely populated area is a direct indication that people have settled in this particular area and perhaps they are carrying out business activities, due to the high demand of the basic commodities like food and clothing. 3) Irrigated vegetation; the light green colour is a strong indication that tries to testify that the land where the image was taken, practise irrigated vegetation. The surrounding climate may not be able to support the natural growth of plants hence it is irrigated. 4) Presences of snow and ice. This is due to the bright colour. 5) Transport facilities. This is shown with the presences of white like lines that join each other. Question 2 Total marks for this question is 50 Find the land cover/feature of circled places the given image with the help of Google Earth. Enter Bongeen, QLD, into Google Earth’s “fly to”. Bongeen is the approximate centre location of the below image. Find and zoom-in to circled spots (A to H) and describe what you found (land cover/ land use) in each spot in one or few words. Remember, land cover on the ground may have changed in the Google Erath mage. Location in image Approximate land cover type according to your own decision (with the help of Google image) A Remnant Vegetation; this is due to the presence of the brown colour on the ground. B Scrub and Sparsely vegetated. C Urban areas. This is due to the presences of buildings. D Urban areas. E Farmland. The farm here has been tilted or ploughed and ready for planting. F Grasslands or irrigation vegetation. G Forest. H Bush or small forest. Question 3 Total marks for this question is 50 Searching satellite data: MODIS satellite images. Answer questions using the information in following web site. http://earthdata.nasa.gov/data/near-real-time-data/rapid-response/modis-subsets 3.1. Go to “USDA Foreign Agricultural Service (FAS)” sub-link. Locate and click on the image window for Mekong River Delta area. What are the different band combinations and image resolutions available? (Image selection 05, writing 05 marks) Band combination True colours; it offers a natural colour rendition, or when it comes close to the images on focus. This means that they are the colour images that can be visible to the human eye. The Normalized Difference Vegetation Index (NDVI); a simple graphical indicator that can be used to analyse remote sensing measurements, typically but not necessarily from a space platform, and assess whether the target being observed contains live green vegetation or not. MODIS band combination (Grauman, 2007 :6) Image resolutions Spectral resolution; Colour images distinguish light of different spectra. Multispectral images resolve even finer differences of spectrum or wavelength than is needed to reproduce colour. That is, multispectral images have higher spectral resolution than normal colour images. Moderate resolutions Spatial resolution; this is the measure of how closely lines can be resolved in an image, putting into considerations the properties of the system. 3.2. Examine all images for 1st Aug to 20th Aug to find the best available image to monitor the Mekong River Delta area (for Image selection 10 marks). Select and copy (and past into your word file) best image of band 721 TERRA with the date at 200km resolution (Image paste into your assignment 05 marks). Date: 2013/214 - 08/02 Pixel size: 2km | 1km | 500m | 250m 3.3. In “AERONET” subset image, one region of South-Asia is intensively covered by subset images. What is this region and why you thing the area is important be closely monitored? (Finding the location 03 marks, writing 07 marks) South western Pakistan Region. Southeast Asia is a region which stands out globally in satellite observations. During the Asian monsoon (July-September), for example, was clearly visible from the Nasa satellite. These observations shows how the region is unique in it sensitivity to rapidly changing emissions of gas phase and aerosol pollutants in Asia. This is an environmental hazard and thus it calls for constant check up. The region also hosts one of the most complex meteorological and observing environments in the world. There are also fears that due to the internal earth movements, the Southern Asia stands at a risk of strong earthquakes resulting from these. The surface of the Earth is broken into large pieces that are slowly shifting. This gradual process is called "plate tectonics." Researchers and mainly the Nasa, have come up with several supporting evidences that show how the continents are going to move hundreds of millions of years into the future. The image below shows how the continents will behave. 3.4. MODIS is extensively monitoring southern Sahara region of Africa. Under which image subset the user finds these images? Write about 100 words on possible applications of these southern Sahara region images. (Image subset name 03, writing 12 marks) NAfrica_2_03 The southern sub Saharan region is an area that is covered by a striking imagery for desert regions. This feature is important in the management of fire. With the desert like features, fire outbreak can be easily contained since the red colour of fire shall be seen on the satellite which will then send a warning of an emerging fire outbreak. The southern Saharan region of Africa is majorly covered with forest. Without the satellite images sending messages of a fire outbreak, it would make it hard to put off the fire which could end up destroying the whole forest. (Trevor, 2007): 34 Question 4 Total marks for this question is 26 Image Resolution 4.1. Name four different types of image resolutions and write 50-60 words on each of them. (10) 1) Spatial resolution; This is the measure of how closely lines can be resolved in an image. Properties of the system are considered in the measure. Computer monitors generally have a spatial resolution of 72 to 100 lines per inch. Also, this resolution determines the clarity of images. Diffraction, aberrations and atmospheric distortion are the main factors that limit spatial resolution in remote sensing. 2) Pixels Resolutions; Pixels are used as a unit of count to determine the quality of imaging. An image of more pixels can therefore have a resolution less than the amount of its pixels which improves on its Visibility. Some conventions cites resolutions as the total number of pixels in the image, this however can be calculated. 3) Spectral resolution; Lights of different spectra are distinguished by the colour images. However, multispectral images have higher spectral resolution than normal colour images. It can be deducted that multispectral images resolve finer differences of spectrum or wavelength than is needed to get another variety of colour. 4) Radiometric resolution Radiometric resolution determines how finely a system can represent or distinguish differences of intensity, and is usually expressed as a number of levels or a number of bits. When the radiometric resolution is high, the subtle differences of intensity get better. Example the effective radiometric resolution is typically limited by the noise level, rather than by the number of bits of representation. 4.2. To produce a land cover map of Northern Territory, Australia, What is the spatial resolution you select and give reasons for your selection? (08) Radiometric resolution. It describes the spatial structure of an image, the radiometric characteristics describe the actual information content in an image. Under the spatial resolution you have selected, what will be the minimum size of the square shape image (number of pixels in x and y) you need for this work? Use Google Earth to measure the length and width of the state to calculate the x and y pixel numbers. (08) 1:25000 Question 5 Total marks for this question is 24 Image Classification 5.1. What is “unsupervised classification”? (06) Classification involves photo interpretation and using computers to measure pixels and conduct quantitative analysis to spectral classes. Unsupervised classification does not require humans to have knowledge about classes, instead they use clustering algorithm to classify an image data. Migrating means clustering classifier (MMC) is the commonly used method of unsupervised classification (Jennifer, 2006:6). It also involves placing objects into clusters or groupings considering all the features that can only be observed. This is because information or statistics about the data, with their respective class-labels are not known. In case the class label of each object is not provided, then it is viewed as the latent variable and its performance of the clustering scheme are evaluated. 5.2. Briefly explain the conditions that unsupervised classification method is useful (06) Used to cluster pixels in a dataset based on statistics only. This is applied when the pixels do not contain any user training, hence unsupervised classification is applied since it does not require the analyser t have background knowledge of the classes. It is also used in agencies that are involved in long term GIS database maintenance. Unsupervised classification is used in GIS database maintenance today because there are systems that are clustering procedures that are extremely fast and require little in the nature of operational parameters. 5.3. Explain main disadvantages of unsupervised classification method (06) Grauman, K. and Darrell, T. (2005): 47 raise the following; The computation time of MLC unsupervised classification is much slower as compared to that of the MLC supervised classification. A lot of time is consumed to run an iteration hence wasting a lot of time that could have been used to run a different iteration. It is not convenient for the non-experts. For the unsupervised classification, the trained samples have to represent the means of the spreads for each class apart from representing the mean vectors. This is done to avoid getting too much unknown classes. It limits the analyst from total control of menu of classes and its identities. This makes it harder to create a specific menu of informational classes. It only considers spectrally homogenous classes within the data. These classes do not correspond to informational categories that are required by the analyst. This situation poses a big problem of matching spectral classes that originate from the classification to informational classes required by the user. The relationships between informational classes and spectral classes are not constant. This situation makes it hard for the relationships of one image to be extended to others. This is so because the spectral properties of specific informational classes changes over a certain period of time. 5.4. “In unsupervised classification, spectral categories and informational categories need to match”. Explain this statement briefly with one or more examples. (06) It entails the groupings of sets into spaces where they cluster according to their partial-match correspondences. Each image is decomposed into a set of local feature descriptors. Spectral clustering on approximate partial-match similarity level is efficient and produces clusters that group distinct object classes. To improve specificity and to develop a predictive classifier that can label unseen images (Damianos, 2004): 4. A new method is developed that will find prototypical examples in each cluster that are more likely to be class inliers, and then use these prototypes to train a predictive model. Question 6 Total marks for this question is 25 GIS and field data 1. “GIS can be well supported by remote sensing data”. Discuss this statement in about 150 words. (10) GIS is a system of hardware and software that is used to store, retrieve, map and make analysis of geographical data. GIS remote sensing is integrated since remote sensing entails the science of obtaining information from a remotely placed platform. Complete integration of remotely sensed and GIS data has been a long-standing problem though, since it involves integration of incompatible file types. However, the need to integrate has improved due to data availability, Development of large-area forest mapping and the growing need for automated mapping (Barbara, 1997: 27). Aerial and satellite images are the main sources of remote sensing data. These data therefore must be rectified before it can be used in GIS, as those data contains some error which may be due to atmospheric disturbance and camera resolution among others. There is a need for contextual GIS which have been achieved mainly by the ever improvements in remote sensing, which has resulted in higher overall map accuracies. 2. For a GIS database with spatial resolution of 30m and covering 300km by 300km area, you are adding a new land cover data layer. What is the suitable satellite image you will select and give reasons briefly? (08) The SPOT 5 satellite image. This is because as it is higher resolution and would let you get a finer extent against the 30m resolution of the Landsat TM. The SPOT 5 however has a small swath area so you have to use more images. Relative spectral response measurements are assumed to be constant for all detectors covered by a common filter and are therefore normalized to unity at peak response. It is also important to consider repetitive data acquisition cycle. The following is a relative spectral response for LANDSAT TM: 3. If the field investigation is costly and will not conduct for above land cover map, how you verify your land cover map accuracy? Explain. (07) According to Zachary, 2012: 35-40, some of the methods used to verify map accuracy are: Statistical methods; in order to compare stream discharge and sediment concentration according to land use type I can conduct an analysis of variance (ANOVA). The performances’ SWAT model can then be accessed in order to predict discharge and sediment concentrations. I will then calculate the coefficient determination and obtain the collinearity between measured and predicted values, which will range from 0 to 1. Laboratory analysis; for a TSS concentration for example, I could determine it using triplicate analyses for known volumes of samples, then finding the mean, which is the TSS concentration. Stream sampling; here, I can collect the onsite data on a biweekly basis in any number of field seasons. A sampling point will act as the outflow point of all the nine streams. http://www.clarku.edu/departments/geography/pdfs/miller%20%26%20rogan%20book%20chpt6.pdf References Barbara, J. (1997) Fuzzy and isodata classification of landform elements from digital terrain data in Pleasant Valley, Wisconsin. University of Wisconsin Press; Damianos K. (2004) Unsupervised Classification via Decision Trees: An Information Theoretic Perspective, Hopkins, Centre for Language and Speech Processing Grauman, K. & Trevor, D. (2007), Unsupervised Learning of Categories from Sets of Partially Matching Image Features, Cambridge, Massachusetts Institute of Technology Press; Grauman, K. and Darrell, T.(2005) the Pyramid Match Kernel: Discriminative Classification with Sets of Image Features China, Beijing; Johns Hopkins University Press. John, R. & Jennifer, M. (2006) Integrating GIS and Remotely Sensed Data for Mapping Forest Disturbance and Change, Clark University Press; Zachary, A. (2012) Assessing land cover map accuracy and performance of hydrological models for small stream catchments using GIS, Ames Iowa, Iowa State University Press; Read More