Combining the FCM Classifier with Various Kernels to Handle Non-linearity of Class Boundaries

This article presents the use of kernel functions in fuzzy classifiers for an efficient land use/land cover mapping. It focuses on handling mixed pixels obtained from a remote sensing image by considering non-linearity between class boundaries. It uses kernel functions combined with the conventional fuzzy c-means (FCM) classifier. Kernel-based fuzzy c-mean classifiers were applied to classify AWiFS and LISS-III images from Resourcesat-1 and Resourcesat-2 satellites. Optimal kernels were obtained from eight single kernel functions. Fractional images generated from high resolution LISS-IV image were used as reference data. Classification accuracy of the FCM classifier increased with 12.93%. Improvement in overall accuracy shows that non-linearity in the dataset was handled adequately. The inverse multiquadratic kernel and the Gaussian kernel with the Euclidean norm were identified as optimal kernels. The study showed that overall classification accuracy of the FCM classifier improved if kernel functions were included.     

Evaluation of fuzzy-based classifiers for cotton crop identification

In this study, an evaluation of fuzzy-based classifiers for specific crop identification using multi-spectral temporal data spanning over one growing season has been carried out. The temporal data sets have been georeferenced with 0.3 pixel rms error. Temporal information of cotton crop has been incorporated through the following five indices: simple ratio (SR), normalized difference vegetation index (NDVI), transformed normalized difference vegetation index (TNDVI), soil-adjusted vegetation index (SAVI) and triangular vegetation index (TVI), to study the effect of indices on classified output. For this purpose, a comparative study between two fuzzy-based soft classification approaches, possibilistic c-means (PCM) and noise classifier (NC), was undertaken. In this study, advanced wide field sensor (AWiFS) data for soft classification and linear imaging self scanner sensor (LISS III) data for soft testing purpose from Resourcesat-1 (IRS-P6) satellite were used. It has been observed that NC fuzzy classifier using TNDVI temporal index – dataset 2, which comprises four temporal images performs better than PCM classifier giving highest fuzzy overall accuracy of 96.03%.     

Mapping a Specific Crop—A Temporal Approach for Sugarcane Ratoon

Mapping a specific crop using single date multi-spectral imagery remains a challenging task because vegetation spectral responses are considerably similar. The use of multi-temporal images helps to discriminate specific crops as the classifier can make use of the uniqueness in the temporal evolution of the spectral responses of the different vegetated classes. However, one major concern in multi-temporal studies is the selection of optimum dates for the discrimination of crops as the use of all available temporal dates can be counterproductive. In this study this concern was addressed by selecting the best 2, 3, 4… combinations dates. This was done by conducting a separability analysis between the spectral response of the class of interest (here, sugarcane-ratoon) and non-interest classes. For this analysis, we used time series LISS-III and AWiFS sensors data that were classified using Possibilistic c-Means (PCM). This fuzzy classifier can extract single class sub-pixel information. The end result of this study was the detection of best (optimum) temporal dates for discriminating a specific crop, sugarcane-ratoon. An accuracy of 92.8 % was achieved for extracting ratoon crop using AWiFS data whereas the optimum temporal LISS-III data provided a least entropy of 0.437. Such information can be used by agricultural department in selecting an optimum number of strategically placed temporal images in the crop growing season for discriminating the specific crop accurately.     

Extracting Regional Pattern of Wheat Sowing Dates Using Multispectral and High Temporal Observations from Indian Geostationary Satellite

Monitoring of Agricultural crops using remote sensing data is an emerging tool in recent years. Spatial determination of sowing date is an important input of any crop model. Geostationary satellite has the capability to provide data at high temporal interval to monitor vegetation throughout the entire growth period. A study was conducted to estimate the sowing date of wheat crop in major wheat growing states viz. Punjab, Haryana, Uttar Pradesh (UP), Madhya Pradesh (MP), Rajasthan and Bihar. Data acquired by Charged Couple Detector (CCD) onboard Indian geostationary satellite INSAT 3A have continental (Asia) coverage at 1 km?×?1 km spatial resolution in optical spectral bands with high temporal frequency. Daily operational Normalized Difference Vegetation Index (NDVI) product from INSAT 3A CCD available through Meteorological and Oceanographic Satellite Data Archival Centre (MOSDAC) was used to estimate sowing date of wheat crop in selected six states. Daily NDVI data acquired from September 1, 2010 to December 31, 2010 were used in this study. A composite of 7 days was prepared for further analysis of temporal profile of NDVI. Spatial wheat crop map derived from AWiFS (56 m) were re-sampled at INSAT 3A CCD parent resolution and applied over each 7 day composite. The characteristic temporal profiles of 7 day NDVI composite was used to determine sowing date. NDVI profile showed decreasing trend during maturity of kharif crop, minimum value after harvest and increasing trend after emergence of wheat crop. A mathematical model was made to capture the persistent positive slope of NDVI profile after an inflection point. The change in behavior of NDVI profile was detected on the basis of change in NDVI threshold of 0.3 and sowing date was estimated for wheat crop in six states. Seven days has been deducted after it reached to threshold value with persistent positive slope to get sowing date. The clear distinction between early sowing and late sowing regions was observed in study area. Variation of sowing date was observed ranging from November 1 to December 20. The estimated sowing date was validated with the reported sowing date for the known wheat crop regions. The RMSD of 3.2 (n?=?45) has been observed for wheat sowing date. This methodology can also be applied over different crops with the availability of crop maps.     

Wheat crop production estimation using satellite data

This paper reports acreage, yield and production forecasting of wheat crop using remote sensing and agrometeorological data for the 1998–99 rabi season. Wheat crop identification and discrimination using Indian Remote Sensing (IRS) ID LISS III satellite data was carried out by supervised maximum likelihood classification. Three types of wheat crop viz. wheat-1 (high vigour-normal sown), wheat-2 (moderate vigour-late sown) and wheat-3 (low vigour-very late sown) have been identified and discriminated from each other. Before final classification of satellite data spectral separability between classes were evaluated. For yield prediction of wheat crop spectral vegetation indices (RVI and NDVI), agrometeorological parameters (ET_max and TD) and historical crop yield (actual yield) trend analysis based linear and multiple linear regression models were developed. The estimated wheat crop area was 75928.0 ha. for the year 1998–99, which sowed ?2.59% underestimation with land record commissioners estimates. The yield prediction through vegetation index based and vegetation index with agrometeorological indices based models were 1753 kg/ha and 1754 kg/ha, respectively and have shown relative deviation of 0.17% and 0.22%, the production estimates from above models when compared with observed production show relative deviation of ?2.4% and ?2.3% underestimations, respectively.     

Land use land cover classification of Orissa using multi-temporal IRS-P6 awifs data: A decision tree approach

An attempt has been made to understand the potential of temporal Advanced Wide Field Sensor (AWiFS) data aboard IRS-P6 (Resourcesat) to generate the land use land cover information along with the net sown area. The temporal data sets were georeferenced, converted to top of atmosphere reflectance and classified using decision tree classifier, See5. Results indicate that the temporal data set could give a better definition of training sites thereby resulting in good overall kappa (kappa = 0.8651) as well as individual classification accuracies. However, co-registration of temporal datasets accuracies also has got a significant influence on the classification accuracy. Temporal variation in cloud infestation and availability of appropriate data sets within the season (before harvest of the crop) has also affected the classification accuracy.     

Temporal MODIS data for identification of wheat crop using noise clustering soft classification approach

In this study, temporal MODIS-Terra MOD13Q1 data have been used for identification of wheat crop uniquely, using the noise clustering (NC) soft classification approach. This research also optimises the selection of date combination and vegetation index for classification of wheat crop. First, a separability analysis is used to optimise the date combination for each case of number of dates and vegetation index. Then, these scenes have undergone for NC soft classification. The resolution parameter (δ) was optimised for the NC classifier and found to be a value of 1.6 × 10⁴ for wheat crop identification. Classified outputs were analysed by receiver operating characteristics (ROC) analysis for sub-pixel detection. Highest area under the ROC curve was found for soil-adjusted vegetation index corresponding to the three different phenological stages data sets. From this study, the data sets corresponding to the Sowing, Flowering and Maturity phenological stages of wheat crop were found more suitable to identify it uniquely.     

Synergistic use of multi-temporal Radarsat SAR and AWiFS data for <Emphasis Type="Italic">Rabi</Emphasis> rice identification

A national level project on kharif rice identification and acreage estimation is being carried out successfully for several states in the country. A similar methodology based on the temporal profile for identification and delineation of various land cover classes has been followed for the Rabi rice acreage estimation. To define rabi rice, rabi season in India starts from November — February to March — June. Though the main growing season is predominantly winter but the uncertainty of getting cloud free data during the season has resulted in the use of microwave data. A feasibility study was taken up for early forecasting of the rabi rice area using microwave data. Hierarchical decision rule classification technique was used for the identification of the different land cover classes. Land preparation, puddling and transplantation were the reasons for the specific backscatter of rice growing areas. The increase or decrease in the SAR backscatter due to progress in the crop phenology or due to delayed sowing respectively forms the basis for identifying the rice areas. In addition the potential of optical data of a later date has been utilized in the form of various indices from bands including MIR to distinctly separate the late sown areas and also the puddled areas from other areas. This study emphasizes the synergistic use of SAR and optical data for delineating the rabi rice areas which is of immense use in giving an early forecast.     

Evaluation of potential usefulness of irs-1c wifs simulated data for dryland rabi sorghum crop discrimination

Two band simulad WiFS data for five dates correspfonding to rabi sorghun growing season of 1993-94 has been generated for Aurangabad district of Maharashtra. Ground truth data has been used for supervised classificatioa of one date raw image and five date NDVI of simulated WiFS data and the results were compared with those derived from single date IRS LISS I data. Analysis of classification accuracies indicate that single date WIFS data gives slightly lower accuracy of 79 per cent against 81 per cent obtained for single date LISS I data. Overall accuracy for 5-date WiFS data is 96 per cent which shows that classification performance of five date WiFS NDVI data is far superior to the single date data of the IRS-IC WiFS as well as the IRS LISS I. The study thus shows the importance of temporal domain of data acquisition in sorghum crop discrimination, Growth profile for sorghum and other crop classes were generated from multidate WiFS derived NDVI data. Differences in growth profiles of sorghum vigour classes as well as amongst different crop types and forests corroborate the premise of better discrimination of crop types and their vigour on multidate remotely sensed data.     

Crop Evapotranspiration Estimation for Wheat (<Emphasis Type="Italic">Triticum aestivum</Emphasis> L.) Using Remote Sensing Data in Semi-Arid Region of Maharashtra

Penman–Monteith method adapted to satellite data was used for the estimation of wheat crop evapotranspiration during the entire growth period using satellite data together with ground meteorological measurements. The IRS-1D/IRS-P6 LISS-III sensor data at 23.5 m spatial resolution for path 096 and row 059 covering the study area were used to derive, albedo, normalized difference vegetation index, leaf area index and crop height and then to estimate wheat crop evapotranspiration referred to as actual evapotranspiration (ETact). The ETact varied from 0.86 to 3.41 mm/day during the crop growth period. These values are on an average 16.40 % lower than wheat crop potential evapotranspiration (ETc) estimated as product of reference crop evapotranspiration estimated by Penman–Monteith method and lysimetric crop coefficient (Kc). The deviation of ETact from ETc is significant, when both the values were compared with t test for paired two sample means. Though the observations on ETact were taken from well maintained unstressed experimental plot of 120 × 120 m size, there was significant deviation. This deviation could be attributed to, the satellite images representing the actual crop evapotranspiration as function crop canopy biophysical parameters, condition of the crop stand, climatic and soil conditions and the microclimate variation over area of one hectare. However, Penman–Monteith method represents a flat rate of specific growth stage of the crop.     

Composite kernels for hyperspectral image classification

This letter presents a framework of composite kernel machines for enhanced classification of hyperspectral images. This novel method exploits the properties of Mercer's kernels to construct a family of composite kernels that easily combine spatial and spectral information. This framework of composite kernels demonstrates: 1) enhanced classification accuracy as compared to traditional approaches that take into account the spectral information only: 2) flexibility to balance between the spatial and spectral information in the classifier; and 3) computational efficiency. In addition, the proposed family of kernel classifiers opens a wide field for future developments in which spatial and spectral information can be easily integrated.     

Entropy-based fuzzy classification parameter optimization using uncertainty variation across spatial resolution

In the past researchers have suggested hard classification approaches for pure pixel remote sensing data and to handle mixed pixels soft classification approaches have been studied for land cover mapping. In this research work, while selecting fuzzy c-means (FCM) as a base soft classifier entropy parameter has been added. For this research work Resourcesat-1 (IRS-P6) datasets from AWIFS, LISSIII and LISS-IV sensors of same date have been used. AWIFS and LISS-III datasets have been used for classification and LISS-III and LISS-IV data were used for reference data generation, respectively. Soft classified outputs from entropy based FCM classifiers for AWIFS and LISS-III datasets have been evaluated using sub-pixel confusion uncertainty matrix (SCM). It has been observed that output from FCM classifier has higher classification accuracy with higher uncertainty but entropy-based classifier with optimum value of regularizing parameter generates classified output with minimum uncertainty.     

Earthquake-induced built-up damage identification using IRS-P6 data: a comparative study using fuzzy-based classifiers

The 8 October 2005 earthquake caused widespread destruction in both the state of Jammu and Kashmir of India and Northern Pakistan. Due to poor accessibility in the hazardous and difficult mountainous terrain, a proper and comprehensive ground-based survey was not possible. However, with the help of remote sensing data and its analysis techniques, it is feasible to undertake both earthquake-related damage identification and assessment. This study attempts to document and identify built-up damaged (BD) areas using spectral indices taking temporal multispectral images from IRS-P6 LISS-IV. Five spectral indices have been used to identify BD areas using supervised possibilistic c-means (PCM) and noise cluster (NC) classifiers, to analyse the satellite data. The result indicates that Class Based Sensor Independent (CBSI) based Transformed Normalized Difference Vegetation Index (TNDVI) temporal indices provide the best results for identifying BD areas, while Simple Ratio (SR) index gives the best results for built-up undamaged area identification. Further, it observed that PCM classifier performed better in comparison to NC classifier.     

Comprehensive evaluation of soil moisture retrieval models under different crop cover types using C-band synthetic aperture radar data

In the present study, random forest regression (RFR), support vector regression (SVR) and artificial neural network regression (ANNR) models were evaluated for the retrieval of soil moisture covered by winter wheat, barley and corn crops. SVR with radial basis function kernel was provided the highest adj. R² (0.95) value for soil moisture retrieval covered by the wheat crop at VV polarization. However, RFR provided the adj. R² (0.94) value for soil moisture retrieval covered by barley crop at VV polarization using Sentinel-1A satellite data. The adj. R² (0.94) values were found for the soil moisture covered by corn crop at VV polarization using RFR, SVR linear and radial basis function kernels. The least performance was reported using ANNR model for almost all the crops under investigation. The soil moisture retrieval outcomes were found better at VV polarization in comparison to VH polarization using three different models.     

Wheat growth profile: satellite monitoring and crop yield modelling

The most important advantage of the low resolution National Oceanic and Atmospheric Administration’s Advanced Very High Resolution Radiometer (NOAA AVHRR) data is its high temporal frequency and high radiometric sensitivity which helps in vegetation detection in the visible and near-infrared spectral regions. In areas where most of the crop cultivation is in large contiguous areas, and if the AVHRR data are selected for time period such that the crop of interest is well discriminated from other crops, these data can be used for monitoring vegetative growth and condition very effectively. The present study deals with the application of AVHRR data for the monitoring of the wheat crop in its seventeen main growing districts of the Rajasthan state. The fourteen date AVHRR data covering the entire growth period have been used to generate the normalized difference vegetation index (NDV1) growth profile for the crop by masking the non-crop pixels following the two-date NDVI change method. The growth profile parameters and other derived parameters, such as post-anthesis senescence rate and areas under the entire growth profile or under selected growth periods have been related to the district average wheat yield through statistical regression models. Various methods adopted for wheat pixels masking have been critically evaluated. It is found that the wheat yield can be predicted well by the area under the profile in different growth periods.     

Wheat crop classification using multidate IRS LISS-I data

Multi temporal dat acquired at different growth stages increases the dimensionality information content and have advantage over single date data for crop classification. Attempt was made to select suitable single date and combination of multidate data for wheat crop classification in Nalanda district of Bihar state where pulses and other crops are also grown in rabi season. Amongst the single date data February data was found to be better for wheat classification in comparison to November. January, March and April data. Combination of first two principal components each derived from IRS LISS-I four band data acquired in January and February was found to be the best set. Wheat classification accuracy achieved was 94.54 percent.     

通过训练样本采样处理改善小宗作物遥感识别精度

训练样本质量是决定农作物遥感识别精度的关键因素,虽然高空间分辨率卫星的发展有效地解决了农作物遥感识别过程中的混合像元问题,但是当区域内不同作物种植面积差异较大时,训练集中不同类别样本数量往往相差较大,这样的不均衡数据集影响分类器的训练,导致少数类别的识别精度不理想。为研究作物遥感识别过程中的不均衡样本问题,本文基于GF-2号卫星数据,首先挖掘了地物的光谱信息、纹理信息,用特征递归消除RFE (Recursive Feature Elimination)方法进行特征优选,然后从数据处理的角度采用了5种采样算法对不均衡训练集进行处理,最后使用采样后的均衡数据集训练分类器,对比数据采样前后决策树与Adaboost(Adaptive Boosting)两种分类器的识别结果,发现:(1)经过采样处理后两种分类算法明显提升了小宗作物的分类精度;(2)经过ADASYS (Adaptive synthetic sampling)采样处理后,分类器性能提升最多,决策树的Kappa系数提高了14.32%,Adaboost的Kappa系数提高了10.23%,达到最高值0.9336;(3)过采样的处理效果优于欠采样,过采样对分类器的性能提升更多。综上所述,选择合适的采样方法和分类方法是提高不均衡数据集遥感分类精度的有效途径。     

Comparison of two forest types using adeos 1-polder data over west coast of India

Parametric kernel-based models for Bi-directional Reflectance Distribution Function (BRDF) of vegetated land surfaces attempt to describe the BRDF as a linear superposition of a set of kernels describing basic angular shapes. The kernels and associated model coefficients are derived as approximations to the underlying theories of radiative transfer as well as geometric-optical scattering within vegetation canopies. In the present study, five kernels viz. Ross-Thick, Li-Sparse, Ross-Thin, Li-Dense and Roujean geometric-optical (GO) kernels along with a purely empirical model were examined for their suitability in describing the angular characteristics of two forest types (semi-evergreen and deciduous) during November 1996-March 1997 using ADEOS-1 POLDER data. The statistically significant kernels and their combinations with respect to red and NIR (near infra-red) wavelengths were determined using correlation analysis. Atmospherically corrected POLDER measured reflectances at red and NIR wavelengths were regressed with different kernels and their combinations. It is found that the best fitting volume and GO kernel combinations depend on forest type, wavelength and date.     

Trends and Change-Point in Satellite Derived Phenology Parameters in Major Wheat Growing Regions of North India During the Last Three Decades

The virtual certainty of the anticipated climate change will continue to raise many questions about its aggregated impact of environmental changes on our regional food security in imminent future. Crop responses to these changes are certain, but its exact characteristics are hardly understood at regional scale due to complex overlapping effects of climate change and anthropogenic manipulation of agro-ecosystem. This study derived phenology of wheat in north India from satellite data and analyzed trends of phenology parameters over last three decades. The most striking change-point period in phenology trends were also derived. The phenology was derived from two sources: (1) STAR-Global vegetation Health Products-NDVI, and (2) GIMMS-NDVI. The results revealed significant earliness in start of growing season (SOS) in Punjab and Haryana while delay was found in Uttar Pradesh (UP). End of the wheat season almost always occurred early, to even those place where SOS was delayed. Length of growing season increased in most of Punjab and northern Haryana whereas its decrease dominated in UP. The early sowing practice of the farmers of the Punjab and Haryana may be one of the adaptation strategies to manage the terminal heat stress in reproductive stage of the crop in the region. The change-point occurred in late 1990s (1998–2000) in Punjab and Haryana, while in eastern UP it was in early 1990s (1990–1995). Despite the difference in temporal aggregation and spatial resolution, both the datasets yielded similar trends, confirming both the robustness of the results and applicability of the datasets over the region. The results demands further research for proper attribution of the effects into its causes and may help devising crop adaption practices to climatic stresses.     

Column-generation kernel nonlocal joint collaborative representation for hyperspectral image classification

We propose a kernel nonlocal joint collaborative representation classification method based on column generation for hyperspectral imagery. The proposed approach first maps the original spectral space to a higher implicit kernel space by directly taking the similarity measures between spectral pixels as a feature, and then utilizes a nonlocal joint collaborative regression model for kernel signal reconstruction and the subsequent pixel classification. We also develop two kinds of specific radial basis function kernels for measuring the similarities. The experimental results indicate that the proposed algorithms obtain a competitive performance and outperform other state-of-the-art regression-based classifiers and the classical support vector machines classifier.