Ecological model basing on process is an important method to explore questions of climate change which have been a hot topics concerned by scientists. However, the accuracy of a model is key to the accuracy of simulation results. As an important process parameter, the accuracy of optimum temperature to the simulation results has noteworthy influence in the model. In this study, MODIS–NDVI from 2001 to 2010, daily mean temperature from 2001 to 2010 and land cover in 2000 were adopted to extract optimum temperature of vegetation growth for various land cover types in China. In combination with the existing researches, the optimum temperature in this study was defined as all the monthly mean temperature within bounds of suitable temperature for vegetation growth, according to that, generally speaking, plant grow well at the range of suitable temperature. Based on the definition, we explored the optimum temperature of different vegetation types from the perspective of land cover. After analyzing the correlation between corresponding NDVI and temperature for different vegetation types, the optimum temperature of different vegetation types, which could be used to advance parameters optimization scheme for ecological process model, were acquired by overlay analysis from NDVI, temperature and land cover based on the above analysis. The results from this study showed that a significant difference in terms of optimum temperature existed in different vegetation types. The optimum temperature of vegetation types mainly distributed in northern China and Qinghai-Tibet Plateau, the value of which are generally lower than 20℃, are lower than that of vegetation types in southern China. The optimum temperature of evergreen coniferous forest, evergreen broad-leaved forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest, shrub, grassland, crop and residents land were 22.4℃, 23.4℃, 14.1℃, 19.5℃, 20.7℃, 22.6℃, 15.4℃, 24.8℃ and 25.6℃ respectively.
Spatialization of attribute data is a way to output grid data products from vector data. It is beneficial to integrated analysis of geosciences data from various sources and in different formats. However, it is also a process companied with errors, and the errors are closely related to density of data sources, spatializing models and resolution of grid cells. In this paper, 7 levels of density of meteorological stations, 5 spatializing models and 19 levels of resolutions of grid cells were used to analyze the relationships between the errors of annual mean air temperature data spatialization and these affecting factors. The following conclusions were drawn: (a) Reduction of density of meteorological stations led to increasing of the spatialization errors. (b) Of the five models, Adjusted IDW, Regression and Anusplin had higher accuracy than IDW and Kriging. The reason is that both IDW and Kriging are spatial autocorrelation based interpolation methods. They neglect influence of underlying surface on air temperature. But, elevation factor is taken into account for Adjusted IDW, Regression and Anusplin. Therefore higher accuracy can be gained with the three interpolation methods. (c) The accuracy generally decreased with increasing of size of grid cells. The trend was significant especially for Adjusted IDW, Regression and Anusplin. (d) Of the three kinds of factors affecting accuracy of spatialization, the models had the greatest impact on the accuracy, the resolution of grid cells second and the density of meteorological stations the lowest. (e) For spatialization products of annual mean air temperature data at national scale, some spatial hetero-correlation interpolation methods, such as Adjusted IDW, Regress and Anusplin should be applied, and the size of grid cells should be smaller than ten kilometers by ten kilometers. In such a case, the mean absolute error for spatialization can be less than one degree centigrade. At last, a quantitative multiple regression model between spatialization errors and the three kinds of affecting factors was established. The model can be used to predict spatialization errors when some of the affecting factors change, so it can provide the basis for drawing up a plan for spatialization of air temperature data.
Panorama can provide 360 degrees view in one hotspot, and it is widely used in street view, digital campus, etc. However, beside the visualization for the scene browsing, these panoramic models are not really three-dimensional model, which can't support precise measurement for the scene from the images. Users generally pay more attention to spatial measurable information of the objects in the field of survey. Therefore, construct one panoramic model which could support measurement would have more extensive prospects. So it is urgent to construct a kind of three-dimensional measurement model for panoramic images. For the aim to realize measurement and analysis on the panorama, this paper proposed a kind of measurable three-dimensional reconstruction approach based on spherical panoramas. We analyzed and constructed the projective geometry relation between two spherical panoramas, and also constructed some orientation and measurement algorithms for spherical panoramic images, thus establishing a data theoretical calculation basis. The experiment shows that the approach is much validated, which could be applied for 3D scene modeling, and something useful is obtained.
The high-speed railway and its surroundings is a complex system of "people, machine and environment". Building the virtual high-speed railway scene can present the railway lines and the terrains around them intuitively, which makes the technicians feel on the spot and allows them to acquire railway attributes and geographic information in a more efficient way so as to support the management decisions and scientific experiments. Linear referencing system is a one-dimensional system. It consists of a set of line features, on which events can be quickly located based on a reference to the line itself rather than through absolute x, y coordinates. So, according to the linear distribution characteristic of the high-speed railway spatial data and the fixed structure properties of feature object, this paper proposes a virtual scene modeling method for high-speed railway based on the linear referencing system. This modeling method abstracts and simplifies the complicated high-speed railway models and transforms them into reusable basic-element models. At the meantime, the corresponding attribute tables are also created. While doing the modeling work, the modeling objects are stored as the event tables in the database and then using linear referencing tools to do spatial locating. Eventually, the virtual high-speed railway scene can be constructed according to the located spatial positions and attitudes. In this paper, we firstly explored the characteristics of high-speed rail line and linear referencing system. Then we discussed some key technologies such as modeling flowchart, model attribute and event table design method, and geometric constraint in detail. Finally, we selected a case study region for carrying out a test. The results show that, this modeling method transforms the objects into event tables to do spatial locating, which reduces the complexity of calculations, and associates the attribute data with certain scene objects. Thus, this modeling method can surely be used in the management decisions as well as scientific experiments.
With the rapid development of cloud technology, "all resources are becoming services" and the "digital earth" is becoming realizable. Although the great progress of observational technology considerably enriched the available data resources, it raised quite a challenge of processing and reprocessing the spatial data of which the utilization ratio was generally low to date. The data processing service in today's cloud technology is an effective strategy for addressing this challenge, but the key solutions lie in how to describe, search and integrate the data processing services and how to find the optimal service in the cloud service pool. The service semantics based on the ontology theory are useful for bridging the knowledge gap between non-professional and professional clients and can help expand the application domains of spatial data processing services. So, this study, by comparing several service matching algorithms in related researches, presents a multi-level algorithm combining the ontology technique for searching the optimal Geoprocess (GP) service. The algorithm introduces the part-of and the sequential relation semantics to describe the parent-child relationships and the predecessor-successor relationships between different services. The multi-level searches are performed by first matching such relational semantics of the services and then executing a conventional service matching algorithm based on the synaptic theory in neural network. Finally, the experiment in this study confirmed the improvement of the algorithm upon the recall and precision ratio.
Accurate temporal and spatial estimation of land surface temperature (LST) is important for evaluating climate change, global hydrological cycle and monitoring urban heat islands (UHI). LSTs with high quality can be routine by using satellite remote sensing. However, characters of both high spatial and temporal resolutions have been difficult. Cloud cover further reduces the useable observations of surface conditions. Monthly LST product (MOD11C3) composited and averaged temperature values at 0.05 degree latitude/longitude grids (CMG) have coarse spatial resolution (~5.5 km). An alternative to the lack of high-resolution observations is to disaggregate LST data using other products of MODIS of 1 km observations. Historically, disaggregation of LST at high resolutions (1 km) has relied on vegetation index, e.g. NDVI (Normalized Difference Vegetation Index). However, this downscaling method is not adequate for areas encompass basin and upland. We applied Digital Elevation Model (DEM), NDVI, Enhanced Vegetation Index (EVI), Albedo, and slope to resolve this drawback by utilizing stepwise regression method with a moving window. The following is the algorithm. Land surface parameter (LSP) data are sampled to the coarser thermal resolution. A stepwise regression is performed between the monthly temperature product and sampled land surface parameters, then a function f (LSP) framed. The parameters of the regression function are applied to LSP data at high, target resolution. Coarse-scale residual field represent variability in temperature driven by other factors other than vegetation and DEM is added back into the high-resolution base map. So, we utilize LSP to sharpen original images. A reasonable rectangle box that making certain pixel be center is outlined for stepwise regression. Function is obtained by stepwise between LST and LSP. Loop and downscale the other pixels until image processed. Coefficients and intercept are saved as images. The disaggregation LST is achieved by substituting images at target resolution to function. The size of the box flowed over the image in this paper is 19 by 19. Stepwise disaggregation algorithm is applied to the resample MOD13A1 and DEM data. The fitting parameters vary with different window scenes. In contrast, the number of DEM entered function is much larger than NDVI. That indicated DEM is more significant than NDVI, EVI, albedo and slope in most fields of the study area, especially in mountain area. The RMSE of downscaling LST is 4.93K. Image sharpening is therefore not a replacement for high-resolution thermal imaging sensors. Nevertheless, in the absence of thermal imagery because of cloud in Sichuan, DisTrad seems to be able to enhance the resolution of MOD11C3 product.
The database construction of resources and environment science of China's surrounding countries is the basic task for resource environment problem research and macro decision-making. The level of database construction of resources and environment science of surrounding areas is lower than that of developed countries because of China's complex geopolitical relationship with other countries. However, the Ministry of Science and Technology, China Academy of Sciences, the National Development Bank and the various ministries have launched resources and environment science database construction work at global, continental or national scale in the past 2-3 years. The national strategy of promoting economic zone of Silk Road and the Maritime Silk Road Construction puts forward higher, more urgent request to the database construction of resources and environment science of surrounding areas. This paper firstly provided an overview of the significant requirement of government and scientific research, summarized the most recent developments and ideas in large scale scientific datasets and platforms formed by USA, European Union, and international organizations. Then, the recent progress of the construction of resources and environmental scientific datasets were comprehensive reviewed and the problems existed were discussed. Finally, the general strategies for establishing resources and environmental database were introduced, including metadata standards design, data quality control, dynamic updating approaches, etc. For the basis and demand of our country, the content of surrounding countries' resources and environment science database should highlight geopolitical issues and macroeconomic policy-making researches, and we should construct a database covering both natural sciences and humanities at multi-scales. On the information processing and service mode, we should use advanced and mature space information technology to enhance the services of the data comprehensively.
The water conservation service of headwaters of the Yangtze River is very important for ecological security in China. It is generally acknowledged that the source of the Yangtze River includes the Tuotuohe headwater, Dangqu headwater and Chumaerhe headwater. Based on the land cover dataset of middle and late 1970s, early 1990s, 2004 and 2008, this paper analyzed temporal and spatial change characteristics of land cover and ecological condition in headwaters of the Yangtze River over the past 30 years, using the method of direction and extent of land cover change, land cover condition index and land cover change index. The results showed that grassland was the main land cover type in this region in 2008, with a ratio of 66.93% in area. The glaciers were mainly distributed in Tuotuohe headwater, while the marshland was mainly distributed in Dangqu headwater. The change rate of land cover condition index during 1970s-1990s, 1990s-2004 and 2004-2008 was -0.15, -0.24 and 0.01, respectively; while the land cover change index of three periods was -0.20, -0.66 and 0.08, respectively. The land cover and ecological condition of the study area had overall experienced a process of degenerated—obviously degenerated—slightly meliorated since 1970s. The mean annual temperature and precipitation from 2004 to 2008 was 0.57℃ and 17.63mm higher than that from 1970s to 2004. Regional climate change over recent years was helpful for natural ecosystem recovery. In addition, implementation of the ecological protection and construction project since 2004 also had certain positive effects on vegetation restoration.
To study the relationship between urban heat island (UHI) effect and urban expansion in Chengdu's main urban area, three Landsat images in the years of 1992, 2001 and 2009 were used to retrieve land surface temperature (LST), built-up land and vegetation coverage of the area. The three thermal images were normalized and scaled to several grades to reduce seasonal difference, then overlaid to produce a difference image by subtracting corresponding pixels in order to find out the change of the UHI among different dates. In the period of the 17-year study, the urban built-up area of Chengdu has increased significantly, which dramatically increased from 91.24 km2 in 1992 to 403.8 km2 in 2009. The extent of the UHI expansion through the study years was due to large scale urban sprawl and the pattern of the UHI has changed from single-center aggregation to polycentric annular distribution. The quantitative analysis of the UHI using Urban-Heat-Island Ratio Index (URI) reveals that the UHI effect in the area has been greatly mitigated in the past 17 years, as the URI has decreased from 0.72 in 1992 to 0.33 in 2009. Regression statistics indicate that the built-up land and vegetation coverage are critical factors for influencing on LST. The built-up land has a positive exponential relationship with LST rather than a simple linear one, which suggests that high percent built-up land could accelerate the rise of LST. The study also demonstrated that the vegetation coverage plays a distinct role on mitigating the UHI effect, which reduces the built-up land while increases vegetation covers, so as to reduce the LST effectively. The increase of vegetation coverage and reasonable planning are beneficial to the UHI mitigation of Chengdu's main urban area.
With the development of urbanization, there is a close relationship between the structure of urban traffic network and urban spatial expansion. Development of traffic system is caused by urban spatial form, land use development, land price and land use layout feature changes, conversely, land use change also acts on the structure of urban traffic network and road density. How to quantitatively describe the strength of urban road network density and urban land-use change is a key scientific issue. This research takes Dongguan City as an example, through remote sensing images in 2001, 2005 and 2009, using the kernel density estimation (KDE) to describe the intensity of newly added urban land use and explore the relationship between urban road network density and urban spatial expansion. The results showed that, during the period of 2001-2005, the city had a great change in road network with the highway as the mainly new road type. Urban expansion was rapid, with a relatively large amount of newly added urban land use area, especially in the edge region where the road network density is high. During the period of 2005-2009, with the development of road network, county road was the mainly new road type, and the newly added urban land use growth also slowed down. The significance test of correlation coefficient showed: during 2001-2009 period in Dongguan City, at the confidence level of 95%, the correlation coefficient R(xy) of the road network density, of the weighted density (x) and of the kernel density (y) are all larger than 0.1946, which indicated that those three factors are significantly related.
Gully erosion is a result of the combined impact of various geomorphological factors. Taking the Zhifanggou small watershed in Ansai County as the study area, six geomorphological factors, such as land use, soil type, length-slope factor (LS), aspect, plan curvature, and topographical wetness index (TWI) have been selected to calculate weights of the various factors. For each factor, we made a thematic map and translated into a grid pattern. Then we calculated the proportion of each class of the factor in gully (DensClas) after reclassifying the factor through ArcGIS. And we calculated gully pixels and total pixels in the study area and gained the ratio value (DensMap), also, the weight (Wi) of factor proportion in gully (DensClas) and gully proportion in the entire study area (DensMap). We can calculate soil erosion susceptibility by the overlay analysis of Wi maps, then classify soil erosion susceptibility into five levels, i.e., very low, low, moderate, high, and very high, and evaluate gully erosion in this article. The results showed: gully erosion occurred easily in the areas which the slope steepness and length is large, and surface humidity is high. Furthermore, gully erosion occurred more easily in the concave of the back where the land use is grass and where covers yellow spongy soils. Considering gully erosion and soil erosion susceptibility, gully erosion occurred easily in areas where soil erosion susceptibility is above moderate, and the proportion up to 90%. High precision and less error shows that our method is universal, which the response accuracy of the weighs from the experimental area to gully erosion in the validation area is 82.43%, not differing very much from the practical value (90.53%). This study will provide a scientific basis to evaluate and control gully erosion, which have an important practical significance.
Fractal Net Evolution Approach (FNEA) is a high precision algorithm for high resolution remote sensing image segmentation. The segmentation algorithm starts with single image objects of one pixel and repeatedly merges them in several loops in pairs to larger units as long as an upper threshold of homogeneity is not exceeded locally. However, it's high computational cost for large images such as the currently high resolution remote sensing images includes two aspects, one is the time-consuming procedure of creation of initial image objects, and the other is the great amount of initial image objects. In order to boost the speed of the algorithm, generally, an effective improvement measure is to select a faster segmentation method to generate the initial image objects, and then carry out the region merging phase. Thus, the main focus of this paper is to tackle this problem by using a parallel process to segment the original image into subsets as the initial image objects firstly. For the original image data preprocessing, a regular data division way is used to divide the original image data into sub-rectangular data blocks, which will be used as the input data and assigned to different threads for the parallel computing. We introduced an improved method of fractal net evolution approach, the main work as follows: (i) Presenting an automatic seed selection method. The initial seeds are automatically selected from the original image data and the seeded image objects are produced where each image object corresponds to a seed. (ii) Proposing a parallel region growing strategy upon the data paralleled segmentation. Moreover, the strategy solves the problem of merging image objects on both sides of the dividing lines as well. (iii) Using the OpenMP parallel technology. Experimental results, including comparison of final segmentation and assessment of computing efficiency, show that the improved method is more effective and the final segmentation result is reproducible. Thus, the generality and reliability of the method proved the practical value of our work.
The normalized difference water index (NDWI), (Green -NIR)/(Green+NIR), proposed by Mcfeeter, is widely used but easily to mix built-up land and water body due to the spectrum similarity on these two bands (green and near infrared reflection) between the two features (water body and built-up land). It is proposed by water indexes such as MNDWI, CIWI and NWI that importing mid-infrared (MIR) band could help solve the problem, as built-up lands have a higher value on MIR compared with NIR. However, more than half of the satellites have not a MIR band, such as Beijing-1 satellite, HJ-1A/B satellites, QuickBird, IKONOS, SPOT1-3 satellites and so on. A false normalized difference water index (FNDWI) has been proposed to fix the problem without access to MIR band. FNDWI replaces the green band in NDWI with a new FGreen (false green) band, which is created by adjusting the original green band with NIR band value. FNDWI has been tested with NDWI on five different typical regions, including urban, suburb, town, village, and non-built-up lands. The experiments reveal that FNDWI has depressed the value of built-up land, highlighted river water body, thus enhanced the differences between water and built-up land by 116% to 335% of NDWI, as well as remained the original NDWI difference between vegetation and water body. Also, it is found that there exits correlation between river width (measured by pixels) and difference enhancement from NDWI to FNDWI. Difference enhancement of thinner river areas is larger than that of wider river areas. Correlation coefficient between river width and difference enhancement reaches -0.82, indicating their apparent negative correlation. In urban, suburb and town regions, water extracting results using NDWI results are polluted by built-up land information while that using FNDWI is fairly clean. Above all, it is concluded that FNDWI is better than NDWI while extracting water bodies around built-up lands, especially on those thin rivers around urban areas.
Information capacity is a quantity unit of pixel density information. Center pixel and neighboring pixels will all be taken into account in the calculation of information capacity. The value of information capacity is closely related to the image gray levels. The more the gray level is, the greater the information capacity value will be. Thus, information capacity can objectively and effectively express land surface spatial structural information. However, the core issue of information capacity theory is the selection of the constraint domain and the determination of parameters. And appropriate setting of parameters is a key technology to ensure the accurateness of information capacity. In this study, 56 different landform areas of Shaanxi Province were selected as test areas, using the research result of remote sensing images in 2007 ETM + and 2008 SPOT5 as experimental data. According to this, two different calculation method of constraints domain in information capacity were adopted by using comparative analysis and mathematical statistics, which analyzed constraint domain selection and spatial distribution of the remote sensing image information capacity. All these experimental results show that information capacity can reflect the surface spatial structure complexity to a certain extent, and it exits a better linear relationship between information capacity and fractal dimension / information entropy, respectively. Information capacity also increases with the increase of fractal dimension and information entropy. Spatial distribution of information capacity is correlative with topographic feature of loess landform, as the same correlation with the surface spatial structure complexity of land cover types in the Central Shaanxi Plain. So, information capacity can be taken as a discriminate factor to identify the surface complexity.
The land cover map of 30m resolution is generated based on object-oriented method using multi-temporal TM/ETM+ images of central-southern part of Zhejiang Province. The technical process is divided into the following steps. First, multi-scale segmentation algorithm using spectral information, texture characteristics and geometric features is employed to the images of the study area, making the object boundary after segmentation consist to the actual terrain boundaries as far as possible. And by establishing the multi-level object feature roles, we can get different types of the land use with their own extraction scales. This paper uses three-layer split system, the first for parent objects such as woodland and farmland, the second for child objects such as evergreen forest and deciduous forest, and the third for smaller objects such as evergreen coniferous forest and deciduous broadleaved forest. Then, through the analysis of these statistic characteristics, attribute characteristics of MNDWI, LBV and morphological characteristics of compactness, aspect ratio which can be used in classification are analysed, and a decision tree model is constructed to implement the 1:2500 00 land cover mapping of the study area. At last, the precision test of the results are made using two methods of visual interpretation and field validation, and the overall accuracy of visual measurement is 87.66% and the precision of field validating is 83.38%. This article focuses on integration of decision tree algorithm, multi-scale segmentation techniques, hierarchical classification and object-oriented classification method. The results show that the classification method based on object-oriented method not only has high precision, but also realizes the boundaries' coinciding of graph spot and practical ground objects and limits the phenomenon of the wrong classification to the mixed pixels very well. It can also eliminate "pepper phenomenon" based on pixel classification.
It has been more than 10 years since Jungar Banner began to apply the policies of forest implement and desertification control. It is important to understand the condition of vegetation recovery, so as to make more proper environment management policies and achieve sustainable economic development in the future. In this paper, we chose Landsat thematic mapper/enhanced thematic mapper plus(TM/ETM+) data in 1990, 2000 and 2011 to derive the vegetation fractional cover in Jungar Banner. And through image analyzing and processing, we got the NDVI values of pure vegetation and pure soil in such three periods. Then we got the vegetation fractional cover distribution maps in 1990, 2000 and 2011 by conducting the pixel dichotomy model. At last we analyzed the temporal and spatial changes of vegetation fractional cover, such as the transfer matrixes, the vegetation restoration/degradation and the driving forces which lead to such changes. Through quantitative analysis, we reached the conclusions that: in the past 21 years, the mean vegetation fractional cover of Jungar Banner has been improved from 15.53% in 1990 to 17.82% in 2000 and to 29.30% in 2011. Most area of Jungar Banner shows the phenomena of vegetation recovery, and the vegetation restoration phenomena is much more obvious from 2000 to 2011, i.e., from 1990 to 2000, 63.06% of the total area shows vegetation restoration phenomena, while from 2000 to 2010, 84.53% of the total area shows vegetation restoration phenomena. The driving forces analysis shows that the changes of vegetation fractional cover in Jungar Banner don't have significant correlation with the rainfall factor, while the changes have a significant correlation with afforestation projects which have been conducted since 2000.
Topographic correction for remotely sensed images is an important preprocessing step to remove the topographic effects in rugged mountainous terrain. In this study, different C-correction strategies (determining the empirical c-parameter for different NDVI intervals, different land use types, and different slope intervals) and scale levels are used to eliminate the effects of topography on Landsat TM images in complex mountains terrain. Performance of the three strategies was tested by visual comparison, correlation analysis between corrected images and the solar illumination angle (cosi), and image classification accuracy. It is attempted to find a C-correction strategy more suitable for mountainous area. The test site selected for this study is Nanting River basin, which is a subbasin of the Nujiang-Salween River. Visual comparisons showed that all the three strategies of C-correction can substantially eliminate negative terrain effects. All the C-correction strategies, similar to the global C-correction, resulted in over-correction phenomenon to different degree. The landuse-specific C-correction performs best on band 1, band 2, band 3 and band 7, the global C correction performs best on the band 4, while the slope-specific C-correction performs best on band 5. To achieve the best effect, different bands can be considered to take different strategies. Although these C-correction strategies can remove negative terrain effects, classification accuracy of Landsat TM images was not improved in our pilot area. Topography obviously affects remote sensing images with high spatial resolution, however, the effects of terrain on remote sensing images with low spatial resolution cannot be ignored.
With the development of high resolution remote sensing images, imaging analysis technology of object-oriented method shows a distinct advantage in the field of information extraction and target recognition. Image segmentation, as a key technology of object-oriented image analysis method, has a vital role to play on the latter feature extraction and application analysis. Watershed transformation is usually adopted for image segmentation because of its unique advantages. However, because of the complexities of high spatial resolution remote sensing image itself, the traditional method of watershed segmentation is difficult to obtain satisfactory results. This paper presents a new multi-scale segmentation method for high resolution remote sensing image based on improved watershed transformation, in order to suppress over-segmentation of watershed transformation, as well as to provide arbitrary-scale segmentation of remote sensing image for object-oriented segmentation method. The algorithm fully considered multi-spectrum, multi-scale and multi-noises characteristics of high spatial resolution remote sensing image. The details are described as follows. Firstly, an anisotropic diffusion filter was used for image smoothing, because this technology can both remove the noises and maintain edges and other important details information of the input image. Secondly, in order to take into account the multi-scale characteristics of remote sensing images, multi-scale morphology gradient was extracted because of its good combination of the advantages of large structural element and small structural element, and then H-minima technology was used to extract tags of gradient image for the latter marker-based watershed algorithm. Finally, an improved fast region-merging algorithm was proposed to achieve the multi-scale segmentation. This paper elaborated the pre-processing filtering, multi-scale gradient, marking extraction and multi-scale region merging aspects, and the experiments showed that the proposed segmentation method effectively restrained over-segmentation of watershed transformation, and had a good performance for segmentation of high spatial resolution remote sensing images.
