Remote Sensing Digital Image Analysis - Research Proposal Example

?Remote Sensing report Insert Remote Sensing report Introduction A vegetation index also called vegetative index (Indices in plural) refers to a single number that enumerates vegetation biomass and/or plant dynamism for every pixel in a remote sensing representation. The representation is calculated using many spectral bands that are responsive to plant biomass and dynamism. The most ordinary vegetation index is the normalized distinction vegetation index (NDVI) which equates the reflectance values of the red and near-infrared areas of the electromagnetic spectrum using formula: NDVI=NIR-RED/NIR+RED In this formula, NIR is the reflectance value of the pixel in the near-infrared band while RED is the reflectance value of the pixel in the red band. The NDVI varies from -1.0 to 1.0 for every pixel in a representation and is vital in discovering regions of ranging levels of plant biomass/dynamism. This means that higher values symbolize high biomass/dynamism. The vegetation indices normally depend on the actuality that green vegetation indicates an authentic combination of low reflectance at perceptible red wavelengths and elevated reflectance at near infrared wavelengths. In similar manner, a ratio of close to infrared (NIR) by red ® light will issue out an approximation (RVI) of the amount of green vegetation present in every pixel. These therefore may be computed as follows: RVI=NIR/R This ratio is well known as the “Ratio Vegetation Index” thus the notion RVI. The ultimate aim of this report is to examine how the ratio vegetation index is computed using Lake Nakuru in Kenya as example. It also looks at the unsupervised and supervised classification and examines their advantages and disadvantages. Part 1: Vegetation Indices Methodology The methodology used in this research is a fieldwork quantitative analysis carried out on the vegetation structure of Lake Nakuru. The ratio vegetation index is calculated where all the steps involved are indicated. Quantitative analysis is one of the best fieldwork methodologies that provide accurate results that are reliable in conclusions and decision-making processes. Analytical Steps The steps for calculating the ration vegetation index are as follows The first step is calculating a ration vegetation index from Landsat TM Data This means that the computed ratio vegetation index (RVI) for the Nakuru Thematic Mapper (TM) representation. For Landsat Thematic Mapper data, the band 3 gauges red, light as TM band 4 gauges near infrared. Lake Nakuru is a little shallow alkaline lake situated south of Nakuru town in Kenya. It is a small lake but it is the world’s most famous place where the greatest bird sight on earth. It is where more than one million pink flamingos gather to the lake to eat the excessive algae, which flourishes in the warm waters of the lake (Jones, Settle and Wyatt, 2006). Scientific scholars have approximated that the population of the flamingos at Lake Nakuru uses about 250, 000 kg of algae per hectare of surface area every year. However, it is not only the flamingos found at Lake Nakuru although they are the most abundant; Pelicans and cormorants are also present in large numbers. In reality, Lake Nakuru is actually a home for more than 400 different species of birds, which means that it is the leading place with variety of bird species than any other place in the country. The second step is adding color to the Ratio Vegetation Index representation (Bonner, Rohde and Miller, 2001). This is done through reloading the Nakuru image although this time one clicks on the raster options tab to add dialog and show as a pseudo color image. Then select raster/Attributes on the viewer menu bar for Nakuru-rvi.img. The following raster editor should be displayed Select edit/colors from the Attribute Editor Menu. The following color Editor should be displayed: From this color tab, try changing different colors at the start and end colors, and choose the Apply button. The Nakuru-rvi.img representation has to change color. Try using start color of brown and an end color of green for best result (Jones, Settle and Wyatt, 2006). The third step is normalizing difference vegetation index (NDVI) where several researches do not employ simple ratio vegetation index, they employ a normalized difference vegetation index (NDVI) as it is considered to reduce the effect of variable look viewpoint and light and scales between familiar limits (-1 and +1) permitting contrast of images from distinct dates. NDVI is normally articulated in the following formula: NDVI = (NIR – R) / (NIR + R) Processing of NOAA Improved Very High Resolution Radiometer (AVHRR) imagery for vegetation checking in Tunisia. The AVHRR data is used to evaluate the relationship between NDVI over the ground green biomass and the percentage vegetation cover. The outcome of the comparison may be either positive or negative. If the outcome of the relationship is positive, then the data could be employed to analyze the difficulties of overgrazing and desertification in Tunisia. Strengths Easy to understand and use Colors can be added to the ratio vegetation index Can be applicable to many other vegetation features (Bonner, Rohde and Miller, 2001) The information can be easily computed to display the vegetation index clearly Weaknesses Must be used by an individual who is versatile with computer usage Involves a lot of processes that are tiresome (Jones, Settle and Wyatt, 2006) The ratio calculations involved are very complicated Results achieved The ratio vegetation index of Lake Nakuru is achieved helps to determine how the bird species benefits from the lake. Discussion Part 2: unsupervised and supervised vegetation In comparison to the a priori use of investigation-provided data in administered classification, unsupervised categorization is a fragmentation of the data space in the nonexistence of any information issued by any analyst. Analyst information is employed only to obtain information division (or ground cover type, or map) stamps to the fragments developed by clustering. With no doubt, this is a merit of the approach. However, it is a time-consuming process calculation by contrasting it to the system for administered classification (Hulchinson, 2000). Imagine a specific classification practice encompasses N phantom bands and C classes. MLC needs CPN (N+1) multiplications where P is the number of pixels in the representation fragment of interest. Through comparison, clustering of the data needs PCI distance gauges for I iterations. Every distance calculation orders N multiplications (normally distance squared is computed preventing the requirement to examine the squire root operation), in order for the total number of duplication for clustering is PCIN. Therefore, the speed contrast of the two approaches is estimated to be (N +1)/I for MLC in comparison to clustering. For instance, for Landsat MSS data, therefore, in a circumstance where all 4 phantom bands are employed, clustering would have to be accomplished within 5 iterations to be speed spirited with MLC (Bonner, Rohde and Miller, 2001). Methodology For the unsupervised classification, the practice is modeled to issue a person with an overview to the unsupervised categorization where the classifiers have been relevant in aloof sensing for several distinct applications, since they have the merit that no previous knowledge is needed to generate a vegetation map. The representation can thus, be automatically split into spectrally different classes. An overview of ISODATA ERDAS IMAGINE employs the ISODATA algorithm to carry out an unsupervised categorization. ISODATA is short form for ‘Iterative Self-Organizing Data Analysis Technique’. It is said to be iterative because it repeatedly performs a whole categorization where it produces a thematic raster layer as well as re-computing statistics. ‘self-Organizing’ is the way in which it discovers the clusters that are intrinsic in the data (Scorer, 2000). The ISODATA clustering technique employs the reduced spectral distance formula to materialize clusters. It starts with either arbitrary clusters means or means of an available signature collection, and every time the clustering occurs again, the means of these clusters are moved (Jones, Settle and Wyatt, 2006). The clusters averages are used for the next iteration. The ISODATA effectiveness repeats the clustering of the representation until either: An utmost number of iterations has taken place or An utmost percentage of unaltered pixels have been achieved between two iterations. The intrinsic file will acquire a gray scale color system if the first cluster means are arbitrary. If the first cluster means are acquired from an available signature collection, then the intrinsic file will employ the colors of this signature collection. You can as well employ the Raster Attribute Editor to alter the color system (Bonner, Rohde and Miller, 2001). Analytical steps Examine the Churn farm data and carry out an unsupervised classification on the farm’s data. The steps are as follows. 1. Click on the classifier idol in the ERDAS IMAGINE icon pane to open this dialog. 2. To carry out an unsupervised classification, click to open the unsupervised classification dialog (in case of supervised classification, you click to open the supervised classification dialog). The efficacy lets you undertake an unsupervised categorization on an img file using the ISODATA algorithm (Hulchinson, 2000). This efficacy can also be retrieved from the signature Editor. The dialog box should be similar to this: After carrying out unsupervised categorization on Churn fam’s data, the next step is to choose processing options. Strengths Once the steps are well followed, the outcome is guaranteed Steps are very clear and easy to follow if keen The process does not require a prior knowledge to create a classification map Weaknesses The process is too complex Hard to understand if not much keen The ISODATA clustering method is tiresome (Hulchinson, 2000) Discussion The vegetation indices (VIs) are blending of surface reflectance at two or more wavelengths modeled to emphasize on a specific vegetation property. They are obtained using the reflectance possessions of vegetation defined in plant foliage. Each of the vegetation indices is modeled to emphasize a specific vegetation property (Richards, 1999). Numerous VIs, approximately 150 have been printed in scientific literature although only a few of them have considerable biophysical foundation or have been methodically tested. ENVI 4.2 issues 27 vegetation indices for detecting the relative profusion of water, pigments and carbon as articulated in the solar-reflected optical spectrum (400 nm to 2500 nm). The choosing process centered on robustness, general applicability and scientific foundation. Each indices category generally issues several techniques to approximate the presence or absence of a single vegetation property. For distinct properties and field circumstances, certain indices within a group issue outcomes with higher validity than others. by doing a comparison of the outcomes of different vegetation indices in a specific group and correlating these to field circumstances gauged on site, you can evaluate the vegetation indices in a specific group performs the job of modeling the unpredictability in your scenes (Scorer, 2000). Through employing the vegetation indices in any group that best designs the gauged field for little measurements, you can considerably raise the quality of the outcomes from any further process. Therefore, vegetation indices is very important in determining variability and process. Results The outcome of utmost likelihood categorization, migrating means clustering categorization of hybrid classification are acceptable and are close and reliable with the initial 3 bands overlay image.  This is because the solid color influences the natural color system in the 3 bands overlay image (Richards, 1999). The merit of classification is clear, it is easy to obtain the physical relevant reflectance or temperature and their multivariate distributes, it is simple to realize the approximate area coverage for each class, which is significant for quantitative analysis. Discussion for supervised and unsupervised classification With supervised classification, examples of the information steams are identified (that is land cover type) for interest in the representation. These are referred to as ‘training sites’. The representation processing software scheme is then employed to establish statistical characterization of the reflectance for every information stream (Richards, 1999). This stage is always referred to as ‘signature analysis’ and may encompass growing a characterization as easy as the mean or the fury of reflectance on every band or as complicated as comprehensive analyses of the average, variances and covariance over entire bands.  After a statistical characterization is obtained for every information class, the representation is then categorized by evaluating the reflectance for every pixel and establishing a conclusion about which of the autographs it resembles most. Unsupervised categorization on the other is a technique that evaluates a big number of unfamiliar pixels and splits into a number of classed centered on native grouping available in the representation values (Scorer, 2000). While supervised categorization requires analyst –precise training data, unsupervised classification does not.  The foundational  principle is that values with a certain cover type need to be close together in the gauging space meaning that they must have same gray levels. At the same time, data in distinct classes need to be relatively well split meaning that they have very distinct gray levels. Bibliography Bonner, W.J.; Rohde, W.G.; Miller, W.A. 2001. Mapping Wildland resources with Digital Landsat and Terrain Data, Remote Sensing and Resource Management; Johannesen, C.J., Sanders, J.L., Eds.; Soil Conservation Society of America: Iowa, IW, USA, pp. 73-80. Hulchinson, C.K. 2000. Techniques for combining landsat and ancillary data for digital classification improvement. Photogramm. Eng. Remote Sens., 48, 123-130. Jones, A.R.; Settle, J.J.; Wyatt, B.K. 2006. Use of digital terrain data in the interpretation of SPOT-1 HRV multispectral imagery. Int. J. Remote Sens, 9, 669-682. Richards, J. A., 1999. Remote sensing digital image analysis: an introduction (second edition) New York Publication. Scorer, R. S., 2000. Cloud reflectance variations in channel-3. Int. J. Remote Sens., 10,675-686. Read More