Remote Sensing and Image Analysis - Assignment Example

Questions and Answers List the main steps in the remote sensing process. The remote sensing process comprises of 4 main steps, which are as follows: a) Categorize the issues “inductive, deductive, and technologic” b) Statistics compilation “In Situ Data” “Remotely Sensed Data” c) Data examination “Analog Image Processing” “Digital Image Processing” “Preprocessing” “Information Enhancement” “Information Extraction” (Campbell) d) Information Output 2. List 5 biophysical variables that can be measured directly using remote sensing a) Planimetric location (x, y) b) Bathymetry c) Foliar water content d) Soil humidity, and e) Surface temperature (Campbell) 3. What is the difference between an active and a passive sensor? Draw a sketch diagram of the main components of each system. “Passive sensors” are premeditated to perceive naturally occurring power. Usually, the SUN is the resource of radiative energy which is either mirrored as it is intended for noticeable wavelengths or engrossed and subsequently “red-emitted,” (Campbell) as it is intended for thermal infrared wavelengths. Passive recognition merely operates when the logically happening power is accessible whereas “Active sensors” utilizes their own energy resource for the object enlightenment. Active sensors offer the potential to attain measurements anytime, despite of the occasion of daylight or time of year. They are possible to be used for probing energy categories that are not adequately illuminated by the sun. (Campbell) Below given is the diagrams for Passive Sensors and Active Sensors respectively. Passive sensor Active sensor 4. Describe the four components of resolution in remote sensing. a) Spatial - It measures the negligible item that is possible to be determined by the sensor on the land characterized by every pixel or network cell present in the picture. (Campbell) b) Spectral – It is the precise wavelengths possible to record by the sensor in the “electromagnetic spectrum.” (Campbell) c) Temporal – It is an explanation of how frequently a sensor acquires images of a meticulous vicinity of interest. For instance, the “Landsat satellite” revisits a region every 16 days while orbiting the earth, whereas the “SPOT satellite” captures a region every 1 -4 days. (Campbell) d) Radiometric – It denotes the figure of feasible brilliance values in every band of statistics and is resolute by the figure of crumb into which the documented energy is separated. (Campbell) 5. What are analogue data? “Analogue data” is an outline of data exhibit in which standards are revealed in realistic form like “curvature.” (Campbell) It is also a type of calculating in which standards are symbolized by unswervingly computable extents like voltages or resistances. 6. How are analogue data converted to digital data? The analogue data is converted to digital data via unremitting sort of electrical signals changeable in “amplitude, frequency or phase” into relative isolated digital signals. (Campbell) 7. What are digital data? The Digital Data is termed as data coded in a distinct and numerical format. (Campbell) 8. Describe how a digital camera captures digital data. 9. What is the resolution (spatial, spectral, temporal and radiometric) of Landsat MSS, TM, NOAA AVHRR, SPOT? Landsat TM - Spatial Resolution: Bands 1-5 and 7 are 30-meter-square pixels; Band 6, the thermal band, acquires 120-meter-square pixels. (Parkinson) Landsat MSS - Spatial Resolution - 30 m, Spectral Resolution - 3 visible, 1 NIR, 2 mid-IR, and Temporal Resolution - 2 weeks (Parkinson) 10. Describe the orbit of Landsat TM. How does this orbit influence the temporal resolution of Landsat data? The Landsat TM courses the globe from the north to south poles and vice versa. Landsat TM finishes an entire course in an approximate of 99 minutes, permitting the satellite to realize over 14 orbits per day. The satellites formulate an absolute exposure of the earth every 16 days. (Parkinson) 11. How does the Landsat scan mirror sensor work? Sensor in Landsat mounts at trimmings of 4 level mirrors that are tilted accordingly to facilitate the 14 bounces for a ray to imitate span of the three mirrors from foundation to correspondent signal. (Parkinson) 12. How does the SPOT push-broom scanner work? Remote sensing systems that build up an image using a linear array of charged coupled devices or CCDs that record each element of a scan line simultaneously without the use of electromechanical components. (Parkinson) Section 2 1. List the main functions used in image processing. The process of image processing consists of mainly 10 functions that are listed below: “Discrete-Time Signals and Systems” (Campbell) “Linear Time-Invariant Systems” “Sampling: Continuous to Discrete” “Digital Filter Design” “Photons to Pixels” “Point-Wise Operations” “Linear Filtering” “Non-Linear Filtering” “Multi-Scale Transforms” “Motion Estimation” (Campbell) 2. Explain how a multi-spectral image can be displayed as a color composite image. Section 3 1. Draw a typical image band histogram with labeled axes. 2. Draw an image band histogram with: single mode; multi mode; low variance; and high variance. 3. Why might a histogram contain more than one mode? 4. List and describe the main univariate statistics used in image processing. Range Population variance Sample variance Standard deviation Skewness 5. List and describe the main multivariate statistics used in image processing. Band Statistics Class Probability Create Signatures Dendrogram Edit Signatures IsoCluster Maximum Likelihood Classification Principal Component Analysis 6. What extra information do multivariate statistics offer over univariate statistics? Section 4 1. Describe line dropout. The circumstance of Line dropout transpires when a digital image appears to hold a black bad scan lines or contains pixel value of zero. (Campbell) 2. Describe band striping. 3. Describe the influence of the atmosphere on remotely sensed data. 4. What is the difference between absolute and relative radiometric atmospheric correction? Section 5 1. List and describe systematic and non-systematic geometric distortions of image data. The “systematic and non-systematic” distortions are geometric distortion that emerges on satellite images and are obligatory to correct prior to accepting. By way of the ground control points (GCPs), the accuracy of geometric alteration is possible to perform. (Campbell) The corrupted-image with blur exposure and blare make mapping amid GCPs and instantaneous areas on the picture complexity. (Campbell)To repeated list the sub-image in the order of GCP from the enhanced picture performing as example image to allocate the location of GCPs for the improvement of satellite images, the connection can also be used in array to gain the accuracy and precision. The geometric alteration process geometrically switches the image coordinates from (x, y) to (y x) where accurate coordinates devoid of geometrical distortion are articulated by (y, x) and input image coordinates by (x, y). (Campbell) 2. Describe spatial interpolation. What methods of spatial interpolation are available? “Spatial interpolation” is exclamation on gathering of more than one variable. The role to be interpolated is acknowledged at known points i.e. and the interruption dilemma comprises of produced values at capricious points i.e. . (Campbell) Methods of Spatial Interpolation: Nearest-neighbor interpolation (regardless of dimension) 2 dimensions Bilinear interpolation Bicubic interpolation Beziers surface Lanczos resampling (Campbell) 3 dimensions Trilinear interpolation Tricubic interpolation (Campbell) 3. Describe intensity interpolation. What methods of intensity interpolation are available? “Intensity interpolation” is a procedure that recuperates the strength information in the breakings of sequential cross sections of 3-d substances. It is very significant in object re-enactment, particularly when the restructured object is to be auxiliary maneuvered for hallucination, itemization or slicing from illogical angles. (Campbell) There are eight methods currently available: a) Fractional contour to contour correspondence establishment b) Displacement field determination, c) Intermediate contour generation, d) Interior rim generation, e) Rim correspondence establishment, f) Line segment correspondence establishment, g) Intensity determination, and h) Gap filling Section 6 3. Why is contrast enhancement needed with most images? Contrast enhancement is a general procedure for enhancing aspect in terrestrial image data. The enrichment depicts delicate patterns already in attendance in the picture by mapping a sort of standards in the contributing picture to a superior range in the displayed image. (Campbell) 3. Use a histogram and look-up table to describe the operation of a min-max contrast stretch. 4. Use a histogram and look-up table to describe the operation of the following contrast stretches: 10% linear; piecewise; histogram equalising; logarithmic; inverse Log. 5. Describe the effect each of these contrast stretches has on a typical image band. Section 7 1. What is the difference between low-pass and high-pass filters? A “low-pass filter” is intended to highlight better, standardized areas of analogous tenor and decrease the minor facet in a picture. (Campbell)Thus, low-pass filters usually serve to level the manifestation of an image. Typical and median filters, repeatedly used for radar metaphors are illustrations of low-pass filters whereas, “High-pass filters” do the converse and provide to sharpen the emergence of fine aspect in an image. One realization of a high-pass filter initially applies a low-pass filter to a picture and subsequently subtracts the consequence from the innovative, leaving behind only the soaring spatial frequency information. (Parkinson) 2. List 3 common low-pass filters. A. Mean B. Median, and C. Mode (Campbell) 3. List 3 common high-pass filters. A. Prewitt Edge Detector B. Sobel Edge Detector, and C. The Difference Metho d. (Campbell) 4. Explain how a mean filter works. In your answer include a typical convolution mask or kernel for this filter. A non-adaptive, low-pass filter where the strength at each test interval in the image is substituted by the mean of pixel values in a moving gap surrounding the sample. A mean filter utilizes a convolution mask of the magnitude “mxn” (Campbell)to consequence in solitary pixel in the yield image. In the above, (x, y) is the location of the output pixel and (s, t) is the location of a pixel in the mask. 4. Explain how a directional filter works. In your answer include a typical convolution mask or kernel for this filter. A frequent utilization of directional filters is as division of a linear combination of distorted basis functions. The deformation is”steered” by a flexible constraint. The state of support of directional basis works in steerable functions is characteristically huge, so a computationally proficient filters which had proved to be practical in such cases as well. (Campbell) Section 8 1. Describe Principal Components Analysis (PCA) using a 2-d scatterplot. Principal Components Analysis is a technique that trims down data dimensionality by performing a covariance analysis among factors. As such, it is appropriate for data sets in various dimensions. An outcome of PCA is envisaged as a 2D or 3D scatter-plot and it reveals mutual relations and extent of relations amid the rudiments spatially. (Parkinson) 2. What is the effect of PCA on an image? The PCA provide a influential tool for information analysis and pattern recognition which is repeatedly used in signal and image processing as a method for data compression, data dimension reduction or their de-correlation as well. (Campbell) 3. What are the main uses of Band Ratios? Band ratios are used to restrain lighting differences attributable to surface “albedo, look angle and topographic effects.” Band ratios toil best when applied to discrete amalgamation features such as the kaolinite (2135 – 2225 nm) doublet Vs “broad Fe absorption” features situated at wavelengths with a reduction of 1000 nm. Band ratios are an outstanding tool for performing introductory evaluations of hyper-spectral data. (Campbell) 4. Explain using a spectral diagram, why vegetation indices often use red and NIR bands. 5. What is the NDVI and how does it work? The “normalized difference of the vegetation index” NDVI is a non-linear alteration of the observable (red) and near-infrared bands of satellite information. (Parkinson) NDVI is defined as the disparity among the perceptible (red) and near-infrared (nir) bands over their summation. The NDVI is a substitute measure of foliage amount and condition. It is linked with foliage awning distinctiveness such as “biomass, leaf area index and percentage of vegetation cover.” (Parkinson) “NDVI = nir - red / nir + red” The NDVI values ranges from -1 to +1 (pixel values 0-255). 6. What is the Tasselled Cap Transformation? The tasselled cap transformation is a helpful tool for constricting spectral data into a small number of bands connected with corporeal scene characteristics. Initially constructed for accepting significant phenomena of harvest improvement in spectral space, the alteration has latent applications in enlightening key jungle characteristics counting species, age and structure. (Parkinson) Section 9 1. What is the difference between supervised and unsupervised classification? Supervised classification relies on the a priori acquaintance of the site and distinctiveness of ground cover variety that are in the image. This is achieved throughout field work, learning of aerial snaps or other sovereign sources of information. (Parkinson) Training areas generally minute and isolated judged against to the bursting image are used to “train” the categorization algorithm to distinguish land cover module based on their shadowy signatures, as found in the picture. Whereas, in unsupervised classification there is Categorization of digital image data by mainframe processing based completely on the photograph data lacking convenience of guidance samples or a-priori acquaintance of the area. The categorization generates ordinary groupings in the picture values called “spectral clusters or classes.” (Campbell) 2. Explain how a training site is used in supervised classification. For classification, the prerequisite for being consistent is less significant, and a training site can be utilized to engender arithmetical parameters for added solitary classes. Once determined of adequate training sites appliance of these parameters using supervised classification approach. (Campbell)It is necessary to perform a predictable supervised classification on image with the training samples. There are four steps involved in the training procedure. Interpretation of Fuzzy Classification Accuracy Assessment Fuzzy C-Means Clustering Supervised Fuzzy C Mean Classification Bibliography Campbell, James B. Introduction to Remote Sensing. Taylor & Francis, 2002. Parkinson, Claire L. Earth from Above: Using Color-coded Satellite Images to Examine the Global Environment. University Science Books, 1997. Read More