An Elevation Based Change Detection using IDLKamal Jain Department of Civil Engineering Indian Institute of Technology (IIT)-Roorkee Roorkee-247667 Uttaranchal, India kjainfce@iitr.ernet.in Abhishek Gupta Department of Civil Engineering Indian Institute of Technology (IIT)-Roorkee Roorkee-247667 Uttaranchal, India abhiguec@iitr.ernet.in Amit Verma Department of Civil Engineering Indian Institute of Technology (IIT)-Roorkee Roorkee-247667 Uttaranchal, India amit8uec@iitr.ernet.in ABSTRACT Change detection is the technique, which is used for the assessment of natural resources, where multi-date images are compared to find out the type and amount of change have occurred. The various applications of change detection are in agricultural, hydrological, forestry, environmental and ecological field. One of the examples is Shifting Cultivation. It is a moving cultivation in which land is left fellow for some period so that land can retain its fertility and this has become one of the main factors for the deforestation in hilly regions. With the help of satellite data one can know the trend and area of shifting cultivation and digital elevation model can help in finding at which height the shifting cultivation is going on. Keeping this in mind, the paper presents the change detection using land use land cover images along with DEM. The Interactive Data Language (IDL) was used for programming and the results were in form of percentage change that occurs in particular class at specified height range. INTRODUCTION Forests, which are the first and foremost occupants of this planet, are one of the most important renewable resources of inestimable value to the people, in a much as they protect and preserve the physical features, prevent floods, and check the flow of sub-soil water and thus help to maintain the productivity of the cultivated land. Forests also supply a variety of essential raw material and minor forest products like, medical plants, fiber resin, dyes, honey, soap nut and horns etc. Forests are the abode of wild life and add to the scenic beauty of the landscape forest is known to play a multiple role in the socio- economic development of a country. Forests occupy about a fifth of the geographical area of India. With the increasing population pressure and demands on lands for development projects, such as those of power, irrigation rail and road communications, industries new housing colonies and townships, there is little space for development of forest area. An expanding rural population is clearing more and more forestland for agriculture. This leads to the continuous qualitative and quantitative loss of our forest resources. Land surveys indicate that the marginal and degraded landmass is almost 40% of the total mass of the country. The local methods of cultivation, like shifting cultivation, also lead to the deterioration of forests. Forest areas, which are damaged due to natural influences like fire, cyclone, flood, disease, insect and pest attack etc, or by human activities like deforestation etc, can be identified using remote sensing imageries of high resolution. Change detection studies of such forests can gives updated information and if the information will be more useful for resource management, if it is related with height. Similarly, the practice of shifting cultivation actually involves the removal and burning of vegetation to create non-permanent clearings which are fallowed to bush or forests for varying lengths of time, but which also includes the temporary removal of vegetation for livelihood. It highly exists in hilly areas. The practice of shifting cultivation is inherently wasteful, as it is not only causes irreparable damage to the plant community, but also creates condition of soil and water erosion through drying up of water resources. It affects the hydrological balance in nature. Change detection can help in finding the trend and area of shifting cultivation and digital elevation model can help in finding at which height the shifting cultivation is going on. There are many other examples like snow cover mapping where change detection in related to height. Keeping this in mind, the idea was to develop the change detection software where height information can be utilized. METHODOLOGY To demonstrate the working of the developed software two different time land use land cover maps and DEM of the same area are required. IDL (the Interactive Data Language) is used to develop change detection software.  Flow chart showing the various steps in the developed software The various steps are as follows: Input data The input data (LULC & DEM) can be in any of the formats like .tiff, .jpg, .bmp, .gif, etc. These input images may be in grey scale or may include a color table which is quantize using COLOR_QUAN routine of IDL and a pseudo-color image is obtained with the corresponding palette to display the image on standard pseudo-color displays. The output image has 256 colors palette. Creating a mask from LULC Masks are used to isolate specific features. A mask is a binary image, made by using relational operators. A binary mask is multiplied by the original image to omit specific areas. Mathematical operations including logic (conditional) operations and statistics were used in coding. Logic operations were used to apply threshold levels to clip the pixel values of an image. The user has to select a particular topological feature / class on the LULC map by giving input ‘pixel range’. The region of interest is separated and is used to create a mask, containing the pixels having a non-zero (one) value corresponding to the selected area on the LULC map. When this mask is multiplied with the LULC, it gives the selected topological feature, the LULC_MASK. Masking of DEM DEM mask is prepared by using LULC mask. This will give height information of clipped LULC map. This produces an image, MASKED_DEM, which can be used for further classification. Classifying the masked image into classes The information obtained above has to be classified for any further usage. Each color on the MASKED_DEM signifies a height range, many colors can be combined to obtain a larger height range. This combining of height ranges can either be done automatically or manually. Auto-Classification (Unsupervised) In this mode, the topological features are classified to pre-defined height ranges corresponding to pixel color values, for example pixel color values from 0 to 50 are classified as one height range, similarly other height ranges are obtained from pixel color values between 51 to 100, 101 to 150, 151 to 200, 201 to 255. Manual Classification (Supervised) In this mode, the user has given an option to categorize a pixel color value to a height range as per his needs. The area occupied by that pixel color in the two maps is then displayed in percentage and also the percentage difference is given. Change detection output The percentage of each pixel color value is calculated using the number of pixels in the masked image and the original image. The process is repeated with the other LULC map. The difference of the two percentages obtained gives the change that has occurred in that topological feature during the time span of the two LULC maps. This data comprising of the two percentages and the percentage difference along with the classification is then stored in a text file. EXPERIMENT & RESULTS The input data for the testing of software were ASTER satellite images Dated: 24 March 2004 and 21 February 2005 Bands: 3 (VNIR) Resolution: 15m Digital Elevation Model (DEM) The study area was Haridwar and its surrounding region, Uttaranchal. The DEM and the two LULC maps are browsed and opened in the software as show in Fig.1. The user then enters a pixel range for the LULC maps, here the range 90-120 is considered, or just by clicking on any of the LULC maps the pixel value will be selected. The color of the pixel and the X and Y coordinates of the pixel will be displayed spontaneously on clicking on the LULC map. On pressing the Process button, the software processes the given images and displays the percentage change at different height interval classes (Auto Classification).  Fig. 1 Showing inputs files Further, the software provides the user with the option of classifying the pixel values to user defined height range manually by selecting a value from the drop down list or giving any other value. For example in the figure the pixel value 5 is selected and assigned a height value of 300-400 m as shown in Fig2. The images corresponding to pixel value 5 are also displayed on the right side.  Fig.2 Showing Manual Classification.  Fig.3 Showing masked LULC and DEM maps. The resulting images are displayed in the ‘Results tab’ as shown in Fig 3. Options are provided to save the displayed imaged and also to view full resolution images.  Fig.4 Output stored in .txt form As shown in fig.4, the output is saved in txt file which give pixel value (or class), percentage of no of pixel in Lulc1 and lulc2, their difference and height information. For example pixel value 5, the number of pixels is 1.330642% in 2004 and 1.314698% in 2005. The percentage change is 0.016175%. Also the height value information is displayed. This means that for the topological feature represented by pixel value range 90 – 120 on the satellite image occupies a percentage area of 1.330642% in 2004 at the height range given by the pixel value 5 on the DEM; the area reduces to 1.314698% in 2005 causing a decrease by 0.016175%. CONCLUSION The developed software is capable of performing change detection using remote sensing data / land use land cover maps, not only this it can relate this change to the height information contained in DEM. This can be very useful in various studies which are carried out in hilly terrain. REFERENCES
|