A Land Use Land Cover Classification System Using Remote .

2y ago
39 Views
2 Downloads
533.02 KB
5 Pages
Last View : 13d ago
Last Download : 2m ago
Upload by : Joanna Keil
Transcription

International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014ISSN 2229-5518515A Land Use Land Cover classification System UsingRemote Sensing data*Nayana S. Ratnaparkhi, **Ajay D. Nagne, ***Dr. Bharti Gawali*Asst. Professor, **Research Student, ***Professor*D.S.M Art’s, Comm. And Science College, Jintur, **, *** Department of Computer Science & Information Technology, Dr. Babasaheb Ambedkar Marathwada UniversityAurangabad, (M.S), INDIA (431004).AbstractstGeospatial Technology has developed at a significant pace over the past two decades and will play a key role in the development of the nations in the 21century. In a developing nation like India where many people reside in the rural area and very few live in urban areas, we require a very structural planning andprocedure such that the developmental activities and infrastructure facilities are available for both urban and rural areas. The rapid expansions of urban areasare due to rise in population, economic growth and migration from rural to urban areas. Therefore serious problem associated with rapid development such asadditional infrastructure, informal settlements, pollution and scarcity of natural resources has to be studied carefully using Geospatial Technology. GIS and RSdata along with collateral data which help in analyze the growth pattern and nature of urban sprawl. In urban expansion, open land is converted into residentialarea, industrial area, transport facility. The accuracy of change-detection system of urban area is important for understanding the relationships and interactionsbetween human and natural phenomenon. In this paper, we investigate the major techniques, among that post-classification comparison and PCA are mostlyused in change detection System.IJSERKeywords: GIS, RS, PCA, CVA, LULC.1.Introduction:Application of Remote Sensing technology have beenidentified and used as an important tool to monitor land useand surface changes. Satellite Remote sensing collects multispectral, multi-resolution, multi-temporal data providing andmonitoring the process of urban land cover changes [1].Changedetection is the process of identifying differences in the state ofan object or phenomenon by observing it at different times [2].In change detection application, it is necessary to use multitemporal datasets to analyze the temporal effects of the objector phenomena [3]. Currently, with increased computercapability and data availability, Remote Sensing [RS] andgeographic information systems [GIS] have become effectivetools for detecting objects and phenomena change [4]. Remotesensing is a process of acquisition of data or information ofobjects or targets, which is located on earth’s surface. For this,sensors are used which are placed on the satellite [5].Remote sensing and geographic system technology ismore useful in management functions and decision supportsystem which are also useful in the planning process ofurbanisation. RS and GIS application can support a varietyrange of planning, analysis and decision support systemsoperations that can make extraordinary effect to thedevelopment and growth of urban area. Instead of findingoptimal solutions for urban problems, bold approaches mustbe developed [6]. Temporal and special reservations allowscientists to monitor and detect changes over a broad scale andhelp planners to obtain or maintain information on variousphenomenasuchas shiftingagriculturepatterns,industrialization, urban expansion, land use land coverchanges [7] [8].Geographic information system is useful tool formeasuring the change between two or more time periods. Ithas the ability to incorporate multi sources of data into achange detection platform, for example the use of multiplelayers, classified images, maps, toposheets provide a greaterability to extract useful information about the change over aparticular area. Moreover, GIS can measure trends in thesechanges by modelling the available data and using statisticaland analytical functions. The benefit of GIS is the provision ofdifferent outputs in different formats, for example the facilityof maps or tables allow users to select appropriate output forextracting the desired information [6] [8] [4]. Many studieshave attempted to use RS and GIS technology for LULCdetection.The article presents many land use land cover changedetection techniques used for study. It also focuses on urbanexpansion of a city with different methodology used indifferent study. Table 1 shows some examples of applicationsthat are investigated using change detection techniques [4].IJSER 2014http://www.ijser.org

516International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014ISSN 2229-5518processing needed before differencing varies with the type ofimage [11].Table 1: Applications with commonly used TechniquesSr.No.Applications1.Land use/land coverchange2.Urban dscape6.Deforestation7.Wetland change2.changeMost commonly usedtechniques.Image differencing, imagerationing, NDVI, CVA, PCA,chi-square,postClassification, hybrid changedetection, ANN, decisiontree, GIS.Image differencing, postclassification, Hybrid changedetection, PCA, GIS, Chisquare, image fusion.NDVI, ANN, CVA, edifferencing,postClassification.Post classification, GIS.Image differencing, also referred as image data, it is asimple technique for implementing and interpreting changedetection. It divides the image pixels into two results: changeor no change. This process is obtained by subtracting a pixel’sdigital number on the image for one date from thecorresponding pixel’s digital number on the image for seconddate. The general process for detecting the change in two datesin image differencing is extracting the change of the image ofdate 2 from the image of date 1. However, the imagedifferencing technique cannot provide sufficient informationabout the change itself. Atmospheric and other non-surfaceradiance characteristics can affect the result of imagedifferencing [12].Image differencing is widely utilised for changedetection in geographical environment [13].2.2 Image ratioingImage rationing is extracting information between tworegistered images from different dates with one or more bandsin an image or rationed, band by band. The data is comparedon a pixel by pixel basis. In image rationing, the unchangedpixel takes the same number for both dates with a grey level.The changed pixel takes a different value and is displayed at alighter or darker level [4].IJSERPost classification, NDVI,image Differencing, PCA.Post-classification, GIS.Land Use and Land Cover TechniquesOne can computeA modern nation, as a modern business, must haveadequate information on many complex interrelated aspects ofits activities in order to make decisions. Land use is only onesuch aspect, but knowledge about land use and land cover hasbecome increasingly important as the Nation plans toovercome the problems of haphazard, uncontrolleddevelopment, deteriorating environmental quality, loss ofprime agricultural lands, destruction of important wetlands,and loss of fish and wildlife habitat. Land use data are neededin the analysis of environmental processes and problems thatmust be understood if living conditions and standards are tobe improved or maintained at current levels [8] [9] [10].One of the prime prerequisites for better use of land isinformation on existing land use patterns and changes in landuse through time [9] [10].Following are the techniques for LULC.2.1 Image DifferencingImage differencing is an image processing techniqueused to determine changes between images. The differencebetween two images is calculated by finding the differencebetween each pixel in each image, and generating an imagebased on the result. For this, the two images must first bealigned so that corresponding points co inside, and theirphotometric values must be made compatible, either by carefulcalibration, or by post-processing. The complexity of pre-Rxkij xkij(t1) / xkij(t2). . . (1)Where, xkij(t2) is the pixel value of bank k for pixel x at row iand column j at time t2. If intensity of reflected energy is nearlythe same in each image than Rxkij 1, this indicates no change.In areas of change the ratio value would be significantlygreater than 1 or less than 1 depending upon the nature ofchanges between two dates [14].2.3 Change Vector Analysis (CVA)Spectral change vector Analysis is based on multitemporal images. Change Vector Analysis can represent boththe direction and magnitude of a change. The magnitude ofchange is determined by constructing a vector in themultispectral feature space. The one end of the vector isspecified by multispectral digital numbers (DN’s) for the firstdate and other end by the DN values for same pixel on seconddate [15]. The magnitude of vector was calculated from theEuclidean Distance between the differences in positions of thesame pixel from different data takes within the spacegenerated by axes. Greenness and Brightness, as follows [16].R (yb-ya)2 (zb-xa)2Where, R Euclidean DistanceIJSER 2014http://www.ijser.orgYa DN values of Greeness from date2yb DN values of Greeness from date1 (2)

517International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014ISSN 2229-5518xa DN values of Brightness from date1xb Dn values of Brightness from date2The decision that a change has occurred is made if themagnitude of the computed spectral change vector exceeds aspecified threshold direction. The direction of vector containsinformation about the type of change. This method wasapplied to forest change detection in northern Idaho and SouthCarolina. For study of Acre state in Brazilian Amazon the CVAtechnique was used to detect and satisfy different types ofchanges in terms of biomass gain and loss. Spectral changevector analysis can avoid not only the classification and timeconsuming effort and accumulated error in the type ofunreasonable defect in post-classification comparison method,but also can use more or even all of the bands to detect changesin pixels and changes in pixel type information [17].2.4 Image FusionImage fusion is a technology that merges two or moreimages from the same area in different sensors andwavelengths. In image fusion, the first step is to prepare theinput images for fusion process. This includes registration andre-sampling of the input images. Registration is to aligncorresponding pixels in the input images. This is usually doneby geo-referencing the images to a map projection such asUTM (Universal Transverse Mercator). If the images are fromdifferent sensors, and even if they are geo-referenced by theimage vendors, a registration process is still necessary toensure that pixels in the input images exactly represent thesame location on the ground [18]. It shows that proposedwavelet transform approach improves the spatial resolution ofa multi-spectral image it also preserve much portion of thespectral component of the image [19].It is currently the most popular method of urbanchange detection. In Post Classification Comparison, each daterectified imagery is independently classified to fit a commonland type schema (equal number and type of land coverclasses). The resulting land cover maps are then overlaid andcompared on a pixel-by-pixel basis. The result is a map of landcover change. This per pixel comparison can also besummarized in a ‘from-to’ change matrix shows every possibleland cover change under the original classification schema andshows the areas of each change class. Post classificationcomparison includes two scheme supervised classification classification is a process when the analysts select a number ofareas for an image and then identifies the type of eachphenomenon on the computer screen. Supervised classificationusually requires training data and prior knowledge of theobjects that are selected for classification. Unsupervisedclassification is a process by which the computer partitions thedata without prior knowledge and then applies thematiclabels. Unsupervised classification usually requires trainingdata and prior knowledge of the objects that are elected forclassification [4].IJSERDetails of cartographic features and interpretabilitycan be realized using multi-sensor image fusion techniques.Enhanced multi sensor data products will prove useful toscientists seeking to maximize the amount of information thatcan be extracted from satellite image data.3.Change Detection Accuracy AssessmentAccuracy assessment was critical for a map generatedfrom any remote sensing data. Error matrix is in the mostcommon way to present the accuracy of the classification result[21]. Overall accuracy, user’s and producer’s accuracies andkappa statistics were then derived from error matrices. Thekappa statistic incorporates the off diagonal elements of theerror matrices and represents agreement obtained afterremoving the proportion of agreement that could be expectedto occur by change [22]. A considerable number of pixels aretaken from classified image and compared with a referencemap of higher authority to evaluate correctness of classificationprocess. The kappa coefficient ranges from 0 to 1, values higherthan 0.7 are considered acceptable, while those equal to orlower than 0.4 identify a very low correlation betweenclassified image and the ground truth [23].4.Evaluation of change Detection TechniquesImage Fusion having five methods for merging theimages. Intensity-Hue-Saturation (HIS), Principal ComponentAnalysis (PCA), High pass filters (HPF), Brovery and wavelettechnique. The fused image outputs were evaluated based onthree characteristics i.e. statically, graphically and bycomparing classification accuracy. Image fusion provides theway to integrate disparate and complementary data to enhancethe information apparent in the images as well as to increasethe reliability of interpretation [20]. Out of all five algorithmswavelet PCA fusion image has high integrated frequencyinformation and has a high certainty in extraction ofconstruction in the study area and it is also found that theunsupervised classification of the comparison of original imageand supervised classification to extract infrastructuralinformation.An analysis of the literature reviewed indicates thatdifferent methods of change detection produce different mapsof cover change. Some change detection techniques, such asimage differencing, image rationing and PCA, do not providesufficient change trend information. These techniques onlyprovide change or no change results, therefore, the trend anddirection of the change is difficult to determine [24]. Postclassification comparison provides more details about theobjects. Some techniques are affected by reducing such errorsto produce high-quality thematic change detection maps. Postclassification comparison always contains omission andcommission errors and needs the selection of a confusionmatrix and its measures to minimize these errors [25].2.5 Post-ClassificationRefer Table 2 : Evaluation of change Detection Techniques.(seeat last)IJSER 2014http://www.ijser.org

518International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014ISSN 2229-5518Table2 : Evaluation of change Detection ingChangeMatrix Image ratioingChange VectorAnalysis (CVA)PrincipalComponentAnalysis(PCA)Post ClassificationComparisonHybrid ChangeDetectionImage Fusion3.ProvidingChangeDirection lliteImagerySatelliteImageryIJSER 5. Data used Conclusion By analyzing the related literature, it isobserved that the selection of appropriate techniquefor detecting change in an object on earth’s surfacedepends on a number of elements, including thecharacteristics of study area, the spatial resolution ofsensor, atmospheric effects, the sun angle etc. Theaccuracy of change detection also depends upon theresolution of spatial and spectral images. Furthermore,the majority of digital change detection techniquesdepend upon the accuracy of geometric registration oftwo images.IJSER 2014http://www.ijser.org

519International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-2014ISSN 2229-5518[14] Ashbinu Singh, Review article digital change detectiontechniques using remotely sensed data, Internationaljournal of remote sensing.References[1]Mapping and Evaluation of urban sprawl using anintegrated approach of Remote sensing and GIStechniques (2012).[15] Abdullah Almutairi and Timonthy A. Warner, changeDetection Accuracy and Image properties: A StudyUsing Simulated Data, Remote Sensing,2010,1508-1529.[2]A. Singh, Review Article Digital Change detectionTechniques using Remotely Sensed Data, Internationaljournal of RS vol. 10 No. 6, 1989, pp 989-1003.[16] Rodrigo Borrego Lorena, Joao Reberto das Santos, YosioEdemir Shimabukuro, Irving Foster Brown andHerrmann Johann Heinrich kux, A change vectorAnalysis technique to monitor LULC in SW BrazilianAmazon: Acrestate.[3]D. Lu. Et. Al, Change detection techniques, Internationaljournal of remote sensing, volume 25, no. 12, 2004, PP2365-2401.[17] Song Xiaolu and Chen Bo, Change Detection usingchange vector Analysis from Landsat TM Images inWuhan, Procedia Environmental sciences 11,238-244.[4]Abdullah F. Al Qurashi and Lalitkumar, A review,investigation the use of remote sensing and GIStechniques to detect LULC change, Advances in RS,2013, 2, 193-204.[5]Emilio Chuvieco and Alfredo Huete, Fundamentals ofsatellite remote sensing, RC Press Tailor and FranacisGroup 2010.[18] Yan Luo, RengLiu and Yu Feng Zhu, Fusion of RemoteSensing Image Base on the PCA Atrous wavelettransform, The international Archives of thephotogrammetry, Remote Sensing and spatialInformation Science Vol XXXVII.imageFusionIJSER[6]Ajay D. Nagne and Dr. Bharti W. Gawali, Transportationnetwork analysis by using remote sensing and GIS Areview, International Journal of Engineering Researchand Application [IJERA], vol.3, issue 3, May –June 2013,PP 070-076.[7]J. Rogan and D.M. Chen, Remote Sensing Technologyfor mapping and monitoring land cover and land usechange, Progress in planning, vol. 61, no.4, 2004, PP 301325.[8]Ajay D. Nagne, Amol D. Vibhute, Bharti W. Gawali andSuresh S. Mehrotra, Spatial Analysis of TransportationNetwork for Town Planning of Aurangabad City byusing Geographic Information System, InternationalJournal of Scientific & Engineering Research, Volume 4,Issue 7, July 2013.[9][19] Manfred Ehlers in multi-sensorstechniques in Remote Sensing.James R Anderson, Ernest E. Hardy, john T, Roach andRichard E. Witmer, A Land Use & land CoverClassification System for use with Remote Sensor Data.[10] Amol D. Vibhute and Dr. Bharti W. Gawali, Analysisand Modelling of Agricultural Land use using RemoteSensing and Geographic Information System: a Review,International Journal of Engineering Research andApplications (IJERA) Vol. 3, Issue 3, May-Jun 2013,pp.081-091.[11] http:/Wikipedia.org/wiki/image-differencing.[12] P. A. Rogerson, Change detection thresholds forremotely sensed images, Journal of Geographicalsystems boi: 10.1007/s101090100076.[20] Jyoti Sarup and Akinchan singhai, Image fusiontechniques for accurate classification of Remote Sensingdata, International Journal of Geomatics andGeosciences Vol.2, No2,2011.[21] Fan F., Weng Q. And Wang Y., Land Use Land Coverchange in Guangzhou, China from 1998 to 2003, basedon Landsat TM/ETM imagery sensor 2007, 7, 13231342.[22] Yuan F., Sawaya K.E., Koeffelholz B.C., and Bauer M.E.,land cover classification and change analysis of twincities(Minnesota) metropolitan areas by multi-temporalLandsat remote Sensing, Remote sensing ofEnvironment 2005,98,317-328.[23] Jambal

2. Land Use and Land Cover Techniques A modern nation, as a modern business, must have adequate information on many complex interrelated aspects of its activities in order to make decisions. Land use is only one such aspect, but knowledge about land use and land cover

Related Documents:

the land use/ land cover maps prepared for the two different years using multi-date satellite data i.e. 2005-06 and 2011-12. The distribution of land use/ land cover classes in the study area in 2005-06 and 2011-12 (Map 3 & 4) is represented in Table 1. Table1: Statistics of land use / land cover

The variety of land use and land cover data needs is exceedingly broad. Current land use and land cov er data are needed for equalization of tax assess ments in many States. Land use and land cover data also are needed by Federal, State, and local agencies for water-resource inventory, flood control, water Cited by: 4883Publish Year: 1976Author: James R. Anderson, Ernest E. Hardy,

2005). On the other hand, land cover as a concept, is used to describe the actual land mass of the area covered by different land use types. And a change in land cover means a change that is brought about by the different land use forms within a geographic space. Two concepts of change occur on space: conversion and modification.

Land use and land cover change has become a central component in current strategies for managing natural resources and monitoring environmental changes. The advancement in the concept of vegetation mapping has greatly increased research on land use land cover chang

land use and land cover changes occurring in Kilifi County in the period between 1990 and 2014; to establish the effects of land use and land cover change on surface runoff and infiltration capacities and to generate

Classification of land use and land cover is an advantage of remote sensing technology which provides all information about land surface. Numerous studies have investigated land cover classification using different broad array of sensors, r

3 Ecosystem carbon stocks and land-use and land-cover change in tropical wetlands 10 3.1 Background 10 3.2 Land-use change 11 3.3 Carbon stock changes associated with land-use changes 12 3.4 Knowledge gaps 12 3.5 Challenges to carbon stock and land-use and land-cover change assessment 13 3.6 Emerging issues and recommendations 14

Alex Rider [5] Anthony Horowitz New York : Speak, 2006. (2011) SUMMARY: Alex Rider, teen spy, has always been told he is the spitting image of the father he never knew. But when he learns that his father may have been an assassin for the most lethal and powerful terrorist organization in the world, Scorpia, Alex's world shatters. Now Scorpia wants him on their side. And Alex no longer has the .