Classification Accuracy Assessment Research Articles

Applying systematic conservation planning principles to palustrine and inland saline wetlands of New Zealand

Summary1. Previous attempts to identify nationally important wetlands for biodiversity in New Zealand were based on expert panel opinions because quantitative approaches were hampered by a lack of data. We apply principles of systematic conservation planning to remote sensing data within a geographical information system (GIS) to identify nationally important palustrine and inland saline wetlands.2. A catchment‐based classification was used to divide New Zealand into 29 biogeographic units. To meet representation goals, all wetland classes need to be protected within each unit.3. We mapped current and historic wetlands down to a minimum size of 0.5 ha. Over 7000 current wetlands were mapped using standardised satellite imagery and a collection of point or polygon data. Historical extent was estimated from soil information refined using a digital elevation model. The current extent of wetlands is 10% of the historic extent, which is consistent with previous estimates.4. A classification was produced using fuzzy expert rules within a GIS to identify seven wetland classes: bog, fen, swamp, marsh, pakihi/gumland, seepage and inland saline. Swamps and pakihi/gumland are the most common, but the former has sustained the greatest reduction in area with only 6% of its historical extent remaining. A preliminary field assessment of classification accuracy in the Otago region found only 60% agreement, mainly because of the misclassification of marshes into swamps.5. Wetland condition was estimated using six measures of human disturbance (natural cover, human‐made impervious cover, introduced fish, woody weeds, artificial drainage and nitrate leaching risk) applied at three spatial scales: the wetland’s catchment, a 30‐m buffer around the wetland and the wetland itself. Measures were transformed and combined into a single condition index. More than 60% of remaining wetlands had condition indices <0.5, probably indicating moderate to severe degradation and loss of native biodiversity.6. Sites were ranked within each biogeographic unit using the wetland classification to ensure a representative set of wetland diversity. Rankings were determined by combining condition and complementarity to calculate conservation effectiveness that was then weighted by irreplaceability. Highest ranked sites in each biogeographic unit were usually the largest remaining wetlands that contained multiple wetland classes. This reflects their potential to protect a diverse range of wetland classes and a high proportion of the remaining habitat.

A hybrid ant colony optimization technique for power signal pattern classification

This paper presents a novel clustering and pattern classification of power signal disturbances using a variant of S-transform, which is termed as a phase corrected wavelet transform. This variant is obtained by taking the inverse Fourier transform of S-transform and is known as time–time transform (TT-transform). The output from the TT-transform based power signal processing is a set of relevant features that is used for visual localization, detection, and disturbance pattern classification. The TT-transform is a method of dividing a primary time series into a set of secondary, time-localized time series, through use of a translatable, scalable Gaussian window. These secondary time series resemble ordinary windowed time series, except that higher frequencies are more strongly concentrated around the midpoint of the Gaussian, as compared with lower frequencies. In this paper the TT-transform is generalized to accommodate arbitrary scalable windows. The generalized TT-transform can be useful in resolving the times of event initiations when used jointly with a related time-frequency distribution, the generalized S-transform.The extracted features are the input to a fuzzy C-means clustering algorithm (FCA) to generate a decision tree for power signal disturbance pattern classification. To improve the pattern classification of the fuzzy C-means decision tree, the cluster centers are updated using a hybrid ant colony optimization technique (HACO). Further a comparative assessment of power signal disturbance pattern classification accuracy for different population based optimization approach like the genetic algorithm (GA) and particle swarm optimization technique are presented in this paper. The various computational simulations presented in this paper reveal significant improvement in the pattern classification accuracy, the average number of function evaluations and processing time, etc.

Remote Sensing and Geospatial Technological Applications for Site-specific Management of Fruit and Nut Crops: A Review

Site-specific crop management (SSCM) is one facet of precision agriculture which is helping increase production with minimal input. It has enhanced the cost-benefit scenario in crop production. Even though the SSCM is very widely used in row crop agriculture like corn, wheat, rice, soybean, etc. it has very little application in cash crops like fruit and nut. The main goal of this review paper was to conduct a comprehensive review of advanced technologies, including geospatial technologies, used in site-specific management of fruit and nut crops. The review explores various remote sensing data from different platforms like satellite, LIDAR, aerial, and field imaging. The study analyzes the use of satellite sensors, such as Quickbird, Landsat, SPOT, and IRS imagery as well as hyperspectral narrow-band remote sensing data in study of fruit and nut crops in blueberry, citrus, peach, apple, etc. The study also explores other geospatial technologies such as GPS, GIS spatial modeling, advanced image processing techniques, and information technology for suitability study, orchard delineation, and classification accuracy assessment. The study also provides an example of a geospatial model developed in ArcGIS ModelBuilder to automate the blueberry production suitability analysis. The GIS spatial model is developed using various crop characteristics such as chilling hours, soil permeability, drainage, and pH, and land cover to determine the best sites for growing blueberry in Georgia, U.S. The study also provides a list of spectral reflectance curves developed for some fruit and nut crops, blueberry, crowberry, redblush citrus, orange, prickly pear, and peach. The study also explains these curves in detail to help researchers choose the image platform, sensor, and spectrum wavelength for various fruit and nut crops SSCM.

Assessing the accuracy of land cover change with imperfect ground reference data

The ground data used as a reference in the validation of land cover change products are often not an ideal gold standard but degraded by error. The effects of ground reference data error on the accuracy of land cover change detection and the accuracy of estimates of the extent of change were evaluated. Twelve data sets were simulated to allow the exploration of the impacts of a spectrum of ground data imperfections on the estimation of the producer's and user's accuracy of change as well as of change extent. Simulated data were used since this ensured that the actual properties of the data were known and to exclude effects due to other sources of ground reference data error; although the impacts of simulated reference data error on two real confusion matrices are also illustrated. The imperfections evaluated ranged from the inclusion of small amounts of known error into the ground reference data through to the extreme situation in which ground data were absent. The results show that even small amounts of error in the ground reference data can introduce large error into studies of land cover change by remote sensing and reinforce the desire to avoid the expression ground truth as this might imply that the data are a gold standard reference. The effect of reference data imperfections was dependent on the degree of association between the errors in the cross-tabulated data sets. For example, in the scenarios investigated, a 10% error in the reference data set introduced an underestimation of the producer's accuracy of 18.5% if the errors were independent but an over-estimation of the producer's accuracy of 12.3% if the errors were correlated. The magnitude of the mis-estimation of the producer's accuracy was also a function of the amount of change and greatest at low levels of change. The amount of land cover change estimated also varied greatly as a function of ground reference data error. Some possible methods to reduce or even remove the impacts of ground reference data error were illustrated. These ranged from simple algebraic means to estimate the actual values of accuracy and change extent if the imperfections were known through to a latent class analysis that allowed the assessment of classification accuracy and estimation of change extent without the use of ground reference data if the underlying model is defined appropriately.

Analysis and Modeling of Urban Land Cover Change in Setúbal and Sesimbra, Portugal

The expansion of cities entails the abandonment of forest and agricultural lands, and these lands’ conversion into urban areas, which results in substantial impacts on ecosystems. Monitoring these changes and planning urban development can be successfully achieved using multitemporal remotely sensed data, spatial metrics, and modeling. In this paper, urban land use change analysis and modeling was carried out for the Concelhos of Setúbal and Sesimbra in Portugal. An existing land cover map for the year 1990, together with two derived land cover maps from multispectral satellite images for the years 2000 and 2006, were utilized using an object-oriented classification approach. Classification accuracy assessment revealed satisfactory results that fulfilled minimum standard accuracy levels. Urban land use dynamics, in terms of both patterns and quantities, were studied using selected landscape metrics and the Shannon Entropy index. Results show that urban areas increased by 91.11% between 1990 and 2006. In contrast, the change was only 6.34% between 2000 and 2006. The entropy value was 0.73 for both municipalities in 1990, indicating a high rate of urban sprawl in the area. In 2006, this value, for both Sesimbra and Setúbal, reached almost 0.90. This is demonstrative of a tendency toward intensive urban sprawl. Urban land use change for the year 2020 was modeled using a Cellular Automata based approach. The predictive power of the model was successfully validated using Kappa variations. Projected land cover changes show a growing tendency in urban land use, which might threaten areas that are currently reserved for natural parks and agricultural lands.

Classification Accuracy of Accelerometer Count Cut Points During Free Play in Children and Adolescents

Previous research has focused on comparing energy expenditure (EE) and accelerometry output in controlled environments in children and adolescents. There are limited data available comparing EE and accelerometers in free-living environments. It is important to determine if accelerometry accurately assesses EE in free-living settings. PURPOSE: The purpose of this study was to determine if Actical accelerometer count cut points accurately classify physical activity intensity levels in children and adolescents during free play. METHODS: Participants (N=22; 17 males and 5 females) ages 5 to 15 yrs old (mean ± sd: age 11 ± 2.9 yrs; BMI 19.5 ± 4.5) were involved in this study. Participants wore a portable metabolic analyzer and Actical accelerometer while engaging in a simulated after-school program visit, either low activity or high activity. The program closely resembled a typical after-school program and consisted of time periods of both structured and unstructured activity. Physical activity (PA) intensity levels (sedentary (SED), light (LPA), moderate (MPA), or vigorous (VPA)) were assessed using activity energy expenditure (AEE) from the portable analyzer and cut points created by Evenson et al. (2008) for the Actical accelerometer. AEE values were classified based on criteria published by Puyau et al. (2004). Sensitivity, specificity, and kappa were calculated to assess classification accuracy. RESULTS: Sensitivity, specificity, and kappa values (respectively) for each PA intensity level follow: SED=0%, 98.6%, 0; LPA=89.7%, 49.4%, 0.21; MPA=9.5%, 92.4%, 0; VPA=63.9%, 87.1%, 0.52. Overall kappa across categories was 0.26. CONCLUSION: The Actical cut points did not accurately classify SED and MOD, since sensitivity was low, specificity was high, and kappa values were zero. Conversely, the Actical cut points were somewhat accurate for LPA and VPA, since sensitivity was fairly high, specificity was high (for VPA), and kappa values were low. The cut points were most accurate for assessing VPA. Previous research in a laboratory-based setting showed comparable sensitivity and specificity to the current values (with the exception of SED), suggesting that classification accuracy may be similar in laboratory versus free play settings. Funded by R01 HD055400 and the NASPEM student grant

Accuracy Assessment for Land Cover Classification Through Topographic Correction and Multi-temporal Images

Accuracy Assessment for Land Cover Classification Through Topographic Correction and Multi-temporal Images

Automatic Spectral Rule-Based Preliminary Classification of Radiometrically Calibrated SPOT-4/-5/IRS, AVHRR/MSG, AATSR, IKONOS/QuickBird/OrbView/GeoEye, and DMC/SPOT-1/-2 Imagery—Part II: Classification Accuracy Assessment

In Part I of this paper, an operational fully automated Landsat-like image spectral rule-based decision-tree classifier (LSRC), suitable for mapping radiometrically calibrated seven-band Landsat-4/-5 Thematic Mapper (TM) and Landsat-7 Enhanced TM+ (ETM+) spaceborne images [eventually synthesized from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Moderate Resolution Imaging Spectroradiometer (MODIS) imaging sensor] into a discrete and finite set of spectral categories, has been downscaled to properly deal with spaceborne multispectral imaging sensors whose spectral resolution overlaps with, but is inferior to Landsat's, namely: 1) Satellite Pour l'Observation de la Terre (SPOT)-4/-5, Indian Remote Sensing Satellite (IRS)-1C/-1D/-P6 Linear Imaging Self-Scanner (LISS)-III, and IRS-P6 Advanced Wide Field Sensor (AWiFS); 2) National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) and Meteosat Second Generation (MSG); 3) Environmental Satellite (ENVISAT) Advanced Along-Track Scanning Radiometer (AATSR); 4) GeoEye-1, IKONOS-2, QuickBird-2, OrbView-3, TopSat, KOrean MultiPurpose SATellite (KOMPSAT)-2, FORMOsa SATellite (FORMOSAT)-2, Advanced Land Observing Satellite (ALOS) Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2), RapidEye, WorldView-2, PLEIADES-1/-2, and SPOT-6/-7; and 5) Disaster Monitoring Constellation (DMC), IRS-P6 LISS-IV, and SPOT-1/-2. LSRC, together with its five downscaled versions, identified, respectively, as the four-band SPOT-like SRC (SSRC), the four-band AVHRR-like SRC (AVSRC), the five-band AATSR-like SRC (AASRC), the four-band IKONOS-like SRC (ISRC), and the three-band DMC-like SRC (DSRC), form the so-called integrated SRC system of systems. In this paper, first, the classification accuracy and robustness to changes in the input data set of SSRC, AVSRC, AASRC, ISRC, and DSRC are assessed, both qualitatively and quantitatively, in comparison with LSRC's. Next, ongoing and future SRC applications are presented and discussed. They encompass: 1) the implementation of operational two-stage stratified hierarchical Remote Sensing (RS) image understanding systems discussed in Part I of this paper; 2) the integration of near real-time satellite mapping services with Internet map servers; and 3) the development of a new approach to semantic querying of large-scale multisensor image databases. These experimental results and application examples prove that the integrated SRC system of systems is operational, namely, it is effective, near real-time, automatic, and robust to changes in the input data set. Therefore, SRC appears eligible for use in operational satellite-based measurement systems such as those envisaged by the ongoing international Global Earth Observation System of Systems (GEOSS) Programme and the Global Monitoring for Environment and Security (GMES) system project.

Mapping Cropland and Major Crop Types across the Great Lakes Basin using MODIS-NDVI Data

This research evaluated the potential for using the MODIS Normalized Difference Vegetation Index (NDVI) 16-day composite (MOD13Q) 250 m time-series data to develop an annual crop type mapping capability throughout the 480,000 km 2 Great Lakes Basin (GLB). An ecoregion-stratified approach was developed using a two-step processing approach that included an initial differentiation of cropland versus non-cropland and subsequent identification of individual crop types. Major crop types were mapped for the calendar years of 2002 and 2007. National Agricultural Statistics Service (NASS) census data were used to assess county level accuracies on a unit area basis (2002), and the NASS Crop Data Layer (C DL ) was used to generate 231,616 reference data points to support a pixel-wise assessment of the MODIS crop type classification (2007) accuracy across the US portion of the CLB. County level comparisons for 2002 indicated 2.2, ―6.8, ―6.0, and ―5.8 percent of area bias errors for corn, soybeans, wheat, and hay, respectively. Detailed pixel-wise accuracy assessments resulted in an overall crop type classification accuracy of 84 percent (Kappa = 0.73) for 2007. Kappa coefficients ranged from 0.74 to 0.69 for individual ecoregions. The user's accuracies for corn, soybean, wheat, and hay were 87, 82, 81, and 70 percent, respectively. There were spatial variations of classification performances across ecoregions, especially for soybean and hay. Field sizes had a direct impact on the variable classification performances across the GLB.

Topographic normalization of satellite imagery for image classification in northeast India

IRS-LISS-III satellite imagery covering Nongkhyllem Wildlife Sanctuary area located within the Ri-Bhoi District of Meghalaya State, northeast India, was used for analysis of the landcover pattern and vegetation types occurring there. A maximum likelihood classifier was used to generate a supervised classification into land-cover types and the vegetation types within the forested area. The preparation of training data sets used thematic maps of the area, and knowledge accruing from extensive personal field visits. Sample field plots were located at 30 different places in the Sanctuary for classification accuracy assessment. The Normalized Difference Vegetation Index (NDVI) was also computed from LISS-III satellite imagery. A digital elevation model (DEM) of the Sanctuary was generated using a GIS. The DEM was used to test the hypothesis that its joint use with the satellite data would increase classification accuracy. This proved to be the case. Bivariate correlation analysis was performed between spectral and DEM variables to cross-check the results. In the example used, that of the rugged terrain in mountainous parts of northeast India, such integration of satellite land-cover data and DEM information appears to be a necessity in improved land-cover mapping for resource planning and utilization.

Consistency of accuracy assessment indices for soft classification: Simulation analysis

Accuracy assessment plays a crucial role in the implementation of soft classification. Even though many indices of accuracy assessment for soft classification have been proposed, the consistencies among these indices are not clear, and the impact of sample size on these consistencies has not been investigated. This paper examines two kinds of indices: map-level indices, including root mean square error ( rmse), kappa, and overall accuracy ( oa) from the sub-pixel confusion matrix (SCM); and category-level indices, including crmse, user accuracy ( ua) and producer accuracy ( pa). A careful simulation was conducted to investigate the consistency of these indices and the effect of sample size. The major findings were as follows: (1) The map-level indices are highly consistent with each other, whereas the category-level indices are not. (2) The consistency among map-level and category-level indices becomes weaker when the sample size decreases. (3) The rmse is more affected by error distribution among classes than are kappa and oa. Based on these results, we recommend that rmse can be used for map-level accuracy due to its simplicity, although kappa and oa may be better alternatives when the sample size is limited because the two indices are affected less by the error distribution among classes. We also suggest that crmse should be provided when map users are not concerned about the error source, whereas ua and pa are more useful when the complete information about different errors is required. The results of this study will be of benefit to the development and application of soft classifiers.

"DATA TEMPERATURE" IN MINIMUM FREE ENERGIES FOR PARAMETER LEARNING OF BAYESIAN NETWORKS

Maximum likelihood method for estimating parameters of Bayesian networks (BNs) is efficient and accurate for large samples. However, the method suffers from overfitting when the sample size is small. Bayesian methods, which are effective to avoid overfitting, present difficulties for determining optimal hyperparameters of prior distributions with good balance between theoretical and practical points of view when no prior knowledge is available. As described in this paper, we propose an alternative estimation method of the parameters on BNs. The method uses a principle, rooted in thermodynamics, of minimizing free energy (MFE). We define internal energies, entropies, and temperature, which constitute free energies. Especially for temperature, we propose a "data temperature" assumption and some explicit models. This approach can treat the maximum likelihood principle and the maximum entropy principle in a unified manner of the MFE principle. For assessments of classification accuracy, our method shows higher accuracy than that obtained using the Bayesian method with normally recommended hyperparameters. Moreover, our method exhibits robustness for the choice of introduced hyperparameters.

Sample size determination for image classification accuracy assessment and comparison

Many factors influence the quality and value of a classification accuracy assessment and evaluation programme. This paper focuses on the size of the testing set(s) used with particular regard to the impacts on accuracy assessment and comparison. Testing set size is important as the use of an inappropriately large or small sample could lead to limited and sometimes erroneous assessments of accuracy and of differences in accuracy. Here, some of the basic statistical principles of sample size determination are outlined, including a discussion of Type II errors and their control. The paper provides a discussion on some of the basic issues of sample size determination for accuracy assessment and includes factors linked to accuracy comparison. With the latter, the researcher should specify the effect size (minimum meaningful difference in accuracy), significance level and power used in an analysis and ideally also fit confidence limits to derived estimates. This will help design a study and aid the use of appropriate sample sizes, as well as facilitate interpretation of results. In particular, it will help avoid problems, such as under-powered analyses, and provide a richer information base for classification evaluation. The paper includes equations that could be used to determine sample sizes for common applications in remote sensing, using both independent and related samples.

Rough set-derived measures in image classification accuracy assessment

Currently, there are two types of measure in image classification accuracy assessment: pixel-level measures and category-level/map-level measures. These have their own limitations for representing the uncertainty at pixel and category/map levels. In addition, some of these measures derived from the error matrix are obtained by collecting reference data and then they may be affected by factors related to the sampling. This paper uses rough set theory to obtain the rough degree, rough entropy, quality of approximation and accuracy of approximation. Incorporating traditional measures, they compose one kind of three-level architecture for the classified image, which contains pixel-level measures, object/category-level measures and map-level measures. Unlike some conventional measures, these new measures can be derived directly from the supervised classification result without collecting reference data. A case study on the Landsat TM image is used to substantiate the conceptual arguments. The results demonstrate that the proposed measures are valid for measuring the accuracy of classified remotely sensed imagery and can provide additional information to conventional measures.

Teaching and Learning Guide for: Application of Virtual Globes in Education

Teaching and Learning Guide for: Application of Virtual Globes in Education

Monitoring of cropland practices for carbon sequestration purposes in north central Montana by Landsat remote sensing

We used an object-oriented approach in conjunction with the Random Forest algorithm to classify agricultural practices, including tillage (till or no-till (NT)), crop intensity, and grassland-based conservation reserve (CR). The object-oriented approach allowed for per-field classifications and the incorporation of contextual elements in addition to spectral features. Random Forest is a classification tree-based advanced classifier that avoids data over-fitting associated with many tree-based models and incorporates an unbiased internal classification accuracy assessment. Landsat satellite imagery was chosen for its continuous coverage, cost effectiveness, and image accessibility. Classification results for 2007 included producer's accuracies of 91% for NT and 31% for tillage when applying Random Forest to image objects generated from a May Landsat image. Low classification accuracies likely were attributed to the misclassification of conservation-based tillage practices as NT. Results showed that the binary separation of tillage from NT management is likely not appropriate due to surface spectral and textural similarities between NT and conservation-type tillage practices. Crop and CR lands resulted in producer's accuracies of 100% and 90%, respectively. Crop and fallow producer's accuracies were 95% and 82% in the 2007 classification, despite post-senesced vegetation; misclassification within the fallow class was attributed to pixel-mixing problems in areas of narrow (< 100 m) strip management. A between-date normalized difference vegetation index approach was successfully used to detect areas having “changed” in vegetation status between the 2007 and prior image dates; classified “changed” objects were then merged with “unchanged” objects to produce crop status maps. Field crop intensity was then determined from the multi-year analysis of generated crop status maps.

A New Method for Accuracy Assessment of Sub-Pixel Classification Results

Problem statement: Classification accuracy assessment is the main stage of information extraction to evaluate the performance of a classifier. According to the output's classifiers (thematic map/fraction map), a properly strategy for accuracy assessment should be taken into consideration. Since pure pixels are used in traditional accuracy assessment of full pixel classifiers they are not suitable for assessment of sub-pixel classifiers. The objectives of this study were to find a standard sub-pixel accuracy assessment method for evaluation of the sub-pixel classifiers. For this purpose many efforts had been taken and recently some methods and measures such as entropy and cross-entropy had been proposed for sub-pixel accuracy assessment. These methods had their own shortcomings which seriously a fuzzy ground truth data set was needed, the matter that is not available simply. Approach: In this study recently sub-pixel classifier accuracy assessment methods were explored and a new method based on correctness coefficient parameter for the sub-pixel accuracy assessment was introduced. In order to evaluate the CC method, a sub-image of the AVIRIS of hyper spectral data was taken over an agricultural area of California, USA in 1994. The study area consisted of 16 classes. Sub-pixel accuracy assessment methods were discussed. The experiment results using AVIRIS data demonstrated the ability of the new accuracy assessment method. Results: Indeed, in proposed method the matching rate of fraction maps with ground truth data was quantified as correctness coefficient parameters. As a result, flexibility and consistency of sub-pixel accuracy assessment certified by correctness coefficient regarding the type of available data and classification methods. The obtained overall CC over LSU method using 120 bands is about 84.9%. In contrast, the obtained results in terms of OA and Kappa coefficient over LSU method which were achieved by maximum value rule on the fraction maps are 86 and 84% respectively. The Kappa coefficient value is close to overall CC of LSU method. Conclusion: Hence, evaluation and experiments demonstrated that the CC method as an accuracy assessment parameter of a soft classifier can be substituted reasonably by traditional accuracy parameters.

Multi-Source Remotely Sensed Data Combination: Projection Transformation Gap-Fill Procedure.

In this work a new gap-fill technique entitled projection transformation has been developed and used for filling missing parts of remotely sensed imagery. In general techniques for filling missing areas of an image break down into three main categories: first multi-source techniques that take advantages of other data sources (e.g. using cloud free images to fabricate the cloudy areas of other images); the second ones that fabricate the gap areas using non-gapped parts of an image itself, this group of techniques are referred to as single-source gap-fill procedures; and the third group which applies methods that are a combination of both mentioned techniques, therefore they are called hybrid gap- fill procedures. Here a new developed multi-source methodology called “projection transformation for filling a simulated gapped area in Landsat7/ETM+ imagery” is introduced. The auxiliary imagery for filling the gaps is an earlier obtained L7/ETM+ imagery. Quality of the technique was evaluated from three points of view: statistical accuracy measuring, visual comparison, and post classification accuracy assessment. These evaluation indicators are compared to the results obtained from a commonly used technique by the USGS, the Local Linear Histogram Matching (LLHM) [1]. Results show the superiority of our technique over LLHM in almost all aspects of accuracy.

Harshness in image classification accuracy assessment

Thematic mapping via a classification analysis is one of the most common applications of remote sensing. The accuracy of image classifications is, however, often viewed negatively. Here, it is suggested that the approach to the evaluation of image classification accuracy typically adopted in remote sensing may often be unfair, commonly being rather harsh and misleading. It is stressed that the widely used target accuracy of 85% can be inappropriate and that the approach to accuracy assessment adopted commonly in remote sensing is pessimistically biased. Moreover, the maps produced by other communities, which are often used unquestioningly, may have a low accuracy if evaluated from the standard perspective adopted in remote sensing. A greater awareness of the problems encountered in accuracy assessment may help ensure that perceptions of classification accuracy are realistic and reduce unfair criticism of thematic maps derived from remote sensing.

Remote sensing imagery in vegetation mapping: a review

Aims Mapping vegetation through remotely sensed images involves various considerations, processes and techniques. Increasing availability of remotely sensed images due to the rapid advancement of remote sensing technology expands the horizon of our choices of imagery sources. Various sources of imagery are known for their differences in spectral, spatial, radioactive and temporal characteristics and thus are suitable for different purposes of vegetation mapping. Generally, it needs to develop a vegetation classification at first for classifying and mapping vegetation cover from remote sensed images either at a community level or species level. Then, correlations of the vegetation types (communities or species) within this classification system with discernible spectral characteristics of remote sensed imagery have to be identified. These spectral classes of the imagery are finally translated into the vegetation types in the image interpretation process, which is also called image processing. This paper presents an overview of how to use remote sensing imagery to classify and map vegetation cover. Methods Specifically, this paper focuses on the comparisons of popular remote sensing sensors, commonly adopted image processing methods and prevailing classification accuracy assessments. Important findings The basic concepts, available imagery sources and classification techniques of remote sensing imagery related to vegetation mapping were introduced, analyzed and compared. The advantages and limitations of using remote sensing imagery for vegetation cover mapping were provided to iterate the importance of thorough understanding of the related concepts and careful design of the technical procedures, which can be utilized to study vegetation cover from remote sensed images.