Fuzzy Classification Algorithm Research Articles

Self-Configuring Hybrid Evolutionary Algorithm for Fuzzy Imbalanced Classification with Adaptive Instance Selection

Abstract A novel approach for instance selection in classification problems is presented. This adaptive instance selection is designed to simultaneously decrease the amount of computation resources required and increase the classification quality achieved. The approach generates new training samples during the evolutionary process and changes the training set for the algorithm. The instance selection is guided by means of changing probabilities, so that the algorithm concentrates on problematic examples which are difficult to classify. The hybrid fuzzy classification algorithm with a self-configuration procedure is used as a problem solver. The classification quality is tested upon 9 problem data sets from the KEEL repository. A special balancing strategy is used in the instance selection approach to improve the classification quality on imbalanced datasets. The results prove the usefulness of the proposed approach as compared with other classification methods.

Multi-sensor analysis of convective activity in central Italy during the HyMeX SOP 1.1

Abstract. A multi-sensor analysis of convective precipitation events that occurred in central Italy in autumn 2012 during the HyMeX (Hydrological cycle in the Mediterranean experiment) Special Observation Period (SOP) 1.1 is presented. Various microphysical properties of liquid and solid hydrometeors are examined to assess their relationship with lightning activity. The instrumentation used consisted of a C-band dual-polarization weather radar, a 2-D video disdrometer, and the LINET lightning network. Results of T-matrix simulation for graupel were used to (i) tune a fuzzy logic hydrometeor classification algorithm based on Liu and Chandrasekar (2000) for the detection of graupel from C-band dual-polarization radar measurements and (ii) to retrieve graupel ice water content. Graupel mass from radar measurements was related to lightning activity. Three significant case studies were analyzed and linear relations between the total mass of graupel and number of LINET strokes were found with different slopes depending on the nature of the convective event (such as updraft strength and freezing level height) and the radar observational geometry. A high coefficient of determination (R2 = 0.856) and a slope in agreement with satellite measurements and model results for one of the case studies (15 October 2012) were found. Results confirm that one of the key features in the electrical charging of convective clouds is the ice content, although it is not the only one. Parameters of the gamma raindrop size distribution measured by a 2-D video disdrometer revealed the transition from a convective to a stratiform regime. The raindrop size spectra measured by a 2-D video disdrometer were used to partition rain into stratiform and convective classes. These results are further analyzed in relation to radar measurements and to the number of strokes. Lightning activity was not always recorded when the precipitation regime was classified as convective rain. High statistical scores were found for relationships relating lightning activity to graupel aloft.

基于自适应提升小波变换和LBP的极光分类算法

本文提出了一种新的基于自适应提升小波变换的双尺度算法、改进的局部二值模式和模糊近邻分类相结合的极光分类算法。该算法在极光图像预处理的基础之上，先是利用自适应提升的小波变换将原始的极光图像分为几个子图像，然后再对各个子图像进行变尺度的高斯滤波。用局部二值模式进行对子图像进行特征的提取，最后用模糊的近邻分类算法对其进行分类。仿真实验证明，首先本文算法的分类效率高于其他极光分类算法，其次就是本文算法对普通的噪声，例如高斯噪声和椒盐噪声，都有较好的鲁棒性。 This paper presents a new dual-scaling algorithm based on adaptive lifting wavelet transform and improved Local Binary Pattern and classification of a combination of fuzzy neighbor Aurora classi-fication algorithm. Based on the aurora image preprocessing, the algorithm is first using adaptive lifting wavelet transform of the original image to divide into several sub-images of Aurora, and then for each sub-image variable scale Gaussian filter, and to conduct sub-picture with the local binary pattern feature extraction, and finally with fuzzy neighbor classification algorithm to classify. Simulation results show that, first, the algorithm classification efficiency is higher than other Aurora classification algorithm, followed by the algorithm for ordinary noise, such as Gaussian noise and salt and pepper noise having better robustness.

Characterization of Mediterranean hail-bearing storms using an operational polarimetric X-band radar

Abstract. This work documents the effective use of X-band radar observations for monitoring severe storms in an operational framework. Two severe hail-bearing Mediterranean storms that occurred in 2013 in southern Italy, flooding two important Sicilian cities, are described in terms of their polarimetric radar signatures and retrieved rainfall fields. The X-band dual-polarization radar operating inside the Catania airport (Sicily, Italy), managed by the Italian Department of Civil Protection, is considered here. A suitable processing is applied to X-band radar measurements. The crucial procedural step relies on the differential phase processing, being preparatory for attenuation correction and rainfall estimation. It is based on an iterative approach that uses a very short-length (1 km) moving window, allowing proper capture of the observed high radial gradients of the differential phase. The parameterization of the attenuation correction algorithm, which uses the reconstructed differential phase shift, is derived from electromagnetic simulations based on 3 years of drop size distribution (DSD) observations collected in Rome (Italy). A fuzzy logic hydrometeor classification algorithm was also adopted to support the analysis of the storm characteristics. The precipitation field amounts were reconstructed using a combined polarimetric rainfall algorithm based on reflectivity and specific differential phase. The first storm was observed on 21 February when a winter convective system that originated in the Tyrrhenian Sea, marginally hit the central-eastern coastline of Sicily, causing a flash flood in Catania. Due to an optimal location (the system is located a few kilometers from the city center), it was possible to retrieve the storm characteristics fairly well, including the amount of rainfall field at the ground. Extemporaneous signal extinction, caused by close-range hail core causing significant differential phase shift in a very short-range path, is documented. The second storm, on 21 August 2013, was a summer mesoscale convective system that originated from a Mediterranean low pressure system lasting a few hours that eventually flooded the city of Syracuse. The undergoing physical process, including the storm dynamics, is inferred by analyzing the vertical sections of the polarimetric radar measurements. The high registered amount of precipitation was fairly well reconstructed, although with a trend toward underestimation at increasing distances. Several episodes of signal extinction were clearly manifested during the mature stage of the observed supercells.

Evaluation and application of hydrometeor classification algorithm outputs inferred from multi‐frequency dual‐polarimetric radar observations collected during HyMeX

A fuzzy logic hydrometeor classification algorithm (HCA), allowing discrimination between six microphysical species regardless of the radar wavelength is presented and evaluated. The proposed method is based upon combination sets of dual‐polarimetric observables (reflectivity at horizontal polarization ZH, differential reflectivity ZDR, specific differential phase KDP, correlation coefficient ρHV) along with temperature data inferred from a numerical weather prediction model output.The performance of the HCA is evaluated using 20 h of multi‐frequency dual‐polarimetric radar data collected during the first Special Observation Period (SOP1) of the Hydrological Cycle in the Mediterranean Experiment (HyMeX). A new method based upon intercomparisons of retrieved hydrometeor data deduced from pairs of neighbouring radars (S‐band vs. S‐band and S‐band vs. C‐band) over a common sampling area is proposed to evaluate the consistency of hydrometor classification outputs. S‐/C‐band radar comparisons generally show better consistency than S‐/S‐band radar comparisons due to issues with the identification of the 0°C isotherm on one of the two S‐band radars. Imperfect attenuation correction at C‐band may also lead into differences in hydrometeor fields retrieved from the C‐ and S‐band radars in convective situations, but retrieved hydrometeor data are globally very consistent from one radar to another. Comparisons against in situ airborne data also confirm the overall good performance of the HCA.In a second experiment, an original method allowing the production of multi‐radar three‐dimensional (3D) hydrometeor fields from single‐radar 2D hydrometeor data is tested on a bow‐echo convective system observed with C‐ and S‐band radars. The resulting 3D hydrometeor fields provide a detailed view of the bow‐echo microphysical structure and confirm the good performance of both the HCA and interpolation technique.

On linking atmospheric circulation patterns to extreme wave events for coastal vulnerability assessments

Atmospheric circulation patterns (CPs) are fundamental drivers of regional wave climates. A fuzzy rule-based classification algorithm has recently been developed to identify these atmospheric features. The algorithm is guided by wave heights and optimises the location, shape and strength of a set of CP classes in order to find features that drive extreme waves. This paper focuses on a method for evaluating the performance of CP classification algorithms and reducing the subjectivity in the selection of classification parameters. We suggest a method based on entropy to quantify the classification quality and provide a means to objectively define an optimal number of CP classes. We also explore the sensitivity to the temporal resolution of the data. For our case study site, the entropy measure indicates that a good quality classification requires 15–20 CP classes. However, regardless of the number of classes used, there is a persistent, common class that explains a large proportion of extreme wave events. The methods described here contribute to developing a new framework for improved statistical wave modelling in coastal vulnerability risk assessments.

A static heuristic approach to detecting malware targets

Nowadays malware writers usually employ several obfuscation techniques to evade detection. The number of variants detected each day has been increasing significantly. Unfortunately traditional detection approaches such as signature scanning are becoming inefficient to detect such malwares. Researches show that these obfuscations make some anomalies in Portable Executable files. In this paper, by focusing on important static heuristic features and fuzzy classification algorithms, we tried to detect malwares and packed files. In addition, we used preprocessing to evade anomaly exceptions in benign files that improved our detection results. The experimental results, using over 63000 file samples, indicate that the proposed detector achieves high detection results with low false positive and false negative rates. Furthermore, our experimental results on new malware samples that had been undetectable for many years by antivirus products and new custom packers, show that our system works well with new and unknown samples too. Copyright © 2015 John Wiley & Sons, Ltd.

Fuzzy logic selection as a new reliable tool to identify molecular grade signatures in breast cancer--the INNODIAG study.

BackgroundPersonalized medicine has become a priority in breast cancer patient management. In addition to the routinely used clinicopathological characteristics, clinicians will have to face an increasing amount of data derived from tumor molecular profiling. The aims of this study were to develop a new gene selection method based on a fuzzy logic selection and classification algorithm, and to validate the gene signatures obtained on breast cancer patient cohorts.MethodsWe analyzed data from four published gene expression datasets for breast carcinomas. We identified the best discriminating genes by comparing molecular expression profiles between histologic grade 1 and 3 tumors for each of the training datasets. The most pertinent probes were selected and used to define fuzzy molecular grade 1-like (good prognosis) and fuzzy molecular grade 3-like (poor prognosis) profiles. To evaluate the prognostic performance of the fuzzy grade signatures in breast cancer tumors, a Kaplan-Meier analysis was conducted to compare the relapse-free survival deduced from histologic grade and fuzzy molecular grade classification.ResultsWe applied the fuzzy logic selection on breast cancer databases and obtained four new gene signatures. Analysis in the training public sets showed good performance of these gene signatures for grade (sensitivity from 90% to 95%, specificity 67% to 93%). To validate these gene signatures, we designed probes on custom microarrays and tested them on 150 invasive breast carcinomas. Good performance was obtained with an error rate of less than 10%. For one gene signature, among 74 histologic grade 3 and 18 grade 1 tumors, 88 cases (96%) were correctly assigned. Interestingly histologic grade 2 tumors (n = 58) were split in these two molecular grade categories.ConclusionWe confirmed the use of fuzzy logic selection as a new tool to identify gene signatures with good reliability and increased classification power. This method based on artificial intelligence algorithms was successfully applied to breast cancers molecular grade classification allowing histologic grade 2 classification into grade 1 and grade 2 like to improve patients prognosis. It opens the way to further development for identification of new biomarker combinations in other applications such as prediction of treatment response.Electronic supplementary materialThe online version of this article (doi:10.1186/s12920-015-0077-1) contains supplementary material, which is available to authorized users.

Fuzzy Logic Classification of S-Band Polarimetric Radar Echoes to Identify Three-Body Scattering and Improve Data Quality

AbstractThe three-body scatter signature (TBSS) is a radar artifact that appears downrange from a high-radar-reflectivity core in a thunderstorm as a result of the presence of hailstones. It is useful to identify the TBSS artifact for quality control of radar data used in numerical weather prediction and quantitative precipitation estimation. Therefore, it is advantageous to develop a method to automatically identify TBSS in radar data for the above applications and to help identify hailstones within thunderstorms. In this study, a fuzzy logic classification algorithm for TBSS identification is developed. Polarimetric radar data collected by the experimental S-band Weather Surveillance Radar-1988 Doppler (WSR-88D) in Norman, Oklahoma (KOUN), are used to develop trapezoidal membership functions for the TBSS class of radar echo within a hydrometeor classification algorithm (HCA). Nearly 3000 radar gates are removed from 50 TBSSs to develop the membership functions from the data statistics. Five variables are investigated for the discrimination of the radar echo: 1) horizontal radar reflectivity factor ZH, 2) differential reflectivity ZDR, 3) copolar cross-correlation coefficient ρhv, 4) along-beam standard deviation of horizontal radar reflectivity factor SD(ZH), and 5) along-beam standard deviation of differential phase SD(ΦDP). These membership functions are added to an HCA to identify TBSSs. Testing is conducted on radar data collected by dual-polarization-upgraded operational WSR-88Ds from multiple severe-weather events, and results show that automatic identification of the TBSS through the enhanced HCA is feasible for operational use.

An accurate fuzzy logic-based fault classification algorithm using voltage and current phase sequence components

In this paper, a fuzzy logic-based fault-type classification algorithm for digital relays is proposed. Using this algorithm, the fault type is recognized amongst 10 types of shunt faults that may occur on transmission lines under different fault resistances, inception angles, line lengths, and system operating conditions. The proposed method uses the phase sequence components of three phase voltages and currents that are available in most of the power system protection relays. An improved technique is used to estimate the current and voltage signals. A fuzzy method is used to identify the type of fault from the current and voltage signals separately and then combines the results to provide more accurate fault-type recognition. This combination in the proposed method leads to overcome the uncertainty associated with each individual signal. In addition, the fuzzy method is rearranged for ease of implementation in the digital relays while maintaining the acceptable results. To test the performance and robustness of the proposed algorithm, various simulation studies have been carried out under wide variations of the parameters. The algorithm is also tested using some real test data recorded from a high-voltage power system. Copyright © 2014 John Wiley & Sons, Ltd.

Pain Expression Recognition Based on Fuzzy Classification Approach

Human activity analysis and expressions recognition play a central role in pervasive health care. Expressions recognition is an active research field in computer vision. In this paper, We propose an approach for automatically recognizing the pain expression from video sequences, in which pain is categorized into four levels, including “no-pain”, “slight pain”, “moderate pain” and “severe pain”. First of all, each frame in video sequences is represented as a “word” by extracting facial velocity information, which is used to charaterize pain expression. Then boosting fuzzy classification algorithm is used for pain expression recognition. The algorithm can automatically categorize and recognize the human pain expression which is contained in the video. Finally we test our algorithm on a pain expression dataset built by ourselves. Experiment results show that the average recognition accuracy is over 91%, which validates its effectiveness. Our results are significantly better than three state-of-the-art approaches for pain expression recognition on the same data.

A New Fuzzy Logic Hydrometeor Classification Scheme Applied to the French X-, C-, and S-Band Polarimetric Radars

AbstractA new fuzzy logic hydrometeor classification algorithm is proposed that takes into account data-based membership functions, measurement conditions, and three-dimensional temperature information provided by a high-resolution nonhydrostatic numerical weather prediction model [the Application of Research to Operations at Mesoscale model (AROME)]. The formulation of the algorithm is unique for X-, C-, and S-band radars and employs wavelength-adapted bivariate membership functions for (ZH, ZDR), (ZH, KDP), and (ZH, ρHV) that were established by using real data collected by the French polarimetric radars and T-matrix simulations. The distortion of membership functions caused by deteriorating measurement conditions (e.g., precipitation-induced attenuation, signal-to-clutter ratio, signal-to-noise ratio, partial beam blocking, and distance) is documented empirically and subsequently parameterized in the algorithm. The result is an increase in the amount of overlapping between the membership functions of the different hydrometeor types. The relative difference between the probability function values of the first and second choice of the hydrometeor classification algorithm is analyzed as a measure of the quality of identification. Semiobjective scores are calculated using an expert-built validation dataset to assess the respective improvements brought by using “richer” temperature information and by using more realistic membership functions. These scores show a significant improvement in the detection of wet snow.

IVTURS: A Linguistic Fuzzy Rule-Based Classification System Based On a New Interval-Valued Fuzzy Reasoning Method With Tuning and Rule Selection

Interval-valued fuzzy sets have been shown to be a useful tool to deal with the ignorance related to the definition of the linguistic labels. Specifically, they have been successfully applied to solve classification problems, performing simple modifications on the fuzzy reasoning method to work with this representation and making the classification based on a single number. In this paper, we present IVTURS, which is a new linguistic fuzzy rule-based classification method based on a new completely interval-valued fuzzy reasoning method. This inference process uses interval-valued restricted equivalence functions to increase the relevance of the rules in which the equivalence of the interval membership degrees of the patterns and the ideal membership degrees is greater, which is a desirable behavior. Furthermore, their parametrized construction allows the computation of the optimal function for each variable to be performed, which could involve a potential improvement in the system's behavior. Additionally, we combine this tuning of the equivalence with rule selection in order to decrease the complexity of the system. In this paper, we name our method IVTURS-FARC, since we use the FARC-HD method to accomplish the fuzzy rule learning process. The experimental study is developed in three steps in order to ascertain the quality of our new proposal. First, we determine both the essential role that interval-valued fuzzy sets play in the method and the need for the rule selection process. Next, we show the improvements achieved by IVTURS-FARC with respect to the tuning of the degree of ignorance when it is applied in both an isolated way and when combined with the tuning of the equivalence. Finally, the significance of IVTURS-FARC is further depicted by means of a comparison by which it is proved to outperform the results of FARC-HD and FURIA, which are two high performing fuzzy classification algorithms.

Research and application of fuzzy tensor machine image classification algorithm

In small sample image classification application,most of traditional classification models take vectors as inputs,which may cause many defects and influence the classification performance.In this paper,the classifier of Fuzzy Support Tensor Machine(FSTM) based on tensor theory and fuzzy support vector machine was proposed.This algorithm took tensors as inputs to obtain the optimal tensor plane.After verifying the performance of the algorithm by using handwritten digital image database,FSTM was applied to triangle node of feather and down category recognition.Compared with the traditional algorithms,FSTM achieves approximately 6.3% increase in correct recognition rate on average.The experimental results show that the FSTM classifier is much more suitable for the application of image classification.

Fuzzy fast classification algorithm with hybrid of ID3 and SVM

The Classification of data is usually very large database that is the reason we want to classify the large data into different fragmentation of its same type. Already many algorithms have been used for classification like Id3, rule based algorithm, decision tree based algorithm, k-nearest-neighbor classification and so on. And these algorithm mainly used for classifying the algorithm accurately and the concept of fast classification is lagging behind in the previous algorithms. In this paper we analysis the efficiency and accuracy of using the entropy, id3 and SVM algorithm with our proposed method of using entropy and fuzzy classification with lower and upper approximation to reduce the computation work for more accuracy classification. We use id3 algorithm to classify the complex member that lie between the lower and upper approximation. Now we use SVM algorithm to classify the other data members thus by hybrid of both the algorithm with our approximation we get the best result of the algorithm Fuzzy Fast Classification FFC. The result of experiments shows that the improved fuzzy fast classification algorithm considerably reduces the computational complexity and improves the speed of classification particularly in the circumstances of the large data.

AN FUZZY NEURAL APPROACH FOR MEDICAL IMAGE RETRIEVAL

Image retrieval based on a query image is necessary for effective and efficient use the information that is stored in medical image databases. Medical Image Retrieval is difficult as not only the localization and directionality of human visual system is to be considered but also the pathological condition. Image identification and segmentation for feature extraction pose a challenge to image retrieval process. Challenges posed include large number of images to be processed for the image retrieval and identifying the region of interest automatically to optimize the search. In this study, we propose a novel image segmentation algorithm Fuzzy Edge Detection and Segmentation (FEDS). The proposed FEDS algorithm is tested on medical images and for classification of images, a bell fuzzy multilayer perceptron is proposed. The proposed neural network Bell Fuzzy Multi Layer Perceptron (BF-MLP) Neural network is constructed by introducing a fuzzy logic in hidden layer with the sugeno model and bell function. The proposed neural network consists of two layers with the first layer being a tanh activation function and the second layer containing the bell fuzzy activation function. The proposed FEDS method was implemented using Matlab and Modelsim. A total of 44 images were considered with three class labels. The edge obtained for which segmentation is done using the proposed segmentation algorithm. The proposed BF-MLP neural network algorithm was implemented using Visual Studio and the classification accuracy compared with MLP Neural Network with sigmoid activation function. In this study, a fuzzy segmentation algorithm and a fuzzy classification algorithm is proposed to improve the medical image retrieval accuracy. The proposed segmentation algorithm, Fuzzy Edge Detection and Segmentation (FEDS), was implemented using Matlab and features were extracted using Fast Hartley Transform (FHT). The extracted features were used to train the proposed neural network, Bell Fuzzy Multi Layer Perceptron Neural Network (BF-MLP). 44 images with 3 class labels were used to test the algorithm and classification accuracy of 93.2% was obtained.

MO‐G‐BRA‐02: Investigation of a Method for Generating Synthetic CT Models from MRI Scans for Radiation Therapy

Purpose: MRI is advantageous for treatment planning, given that MRI can support dose calculations and image‐guided radiotherapy. This study investigates initial methods to derive synthetic CT images from MRI scans that support these needs. Methods: The framework involves a) acquiring multiple MRI image volumes each presenting different tissue/material contrasts, and b) implementing a classification scheme that identifies these materials with sufficient accuracy to assign relative attenuation properties for generating CT images. Six images, consisting of two echoes from ultra‐short TE acquisition (0.07 and 3.7 ms echoes to image cortical bone), T1 GRE, T2 FSE, and fat and water images (from a two‐point Dixon technique), were tested as inputs. Signal intensities of manually‐defined tissues were measured, and used to determine the ability of various image combinations to uniquely classify tissue types. To classify tissue using these images, a fuzzy classification algorithm was used to define membership functions of material types. Assigning attenuation properties to classified tissues yielded synthetic CT image volumes. Results: While subsets of T1, T2, and water/fat image volumes demonstrated statistical significance in identifying fat, solid tissue, and water, the 0.07 ms TE data (UTE1) was necessary to significantly separate bone and air from these tissues. The classification algorithm was able to robustly extract the various material types using subsets of three of the image volumes (T1/UTE1/water). Example CT images and resulting digitally reconstructed radiographs were generated based on attenuation properties applied to the tissue segments. Conclusions: These investigations demonstrated that MRI presents sufficient differential contrasts to separate tissues of interest, and identified the minimum necessary image contrasts for synthetic CT generation using robust classification methods. Further investigations using MRI and CT on subjects undergoing radiotherapy, as well as phantoms, will validate the accuracy of this technique.

Surface texture analysis of a high-resolution DTM: Interpreting an alpine basin

This study focuses on a geomorphometric analysis of a small headwater basin of the Dolomites (Eastern Italian Alps). The analysis of the fine-scale surface morphology was performed on a high-resolution digital elevation model (cell size 2m) derived from an airborne LiDAR survey. The topography was synthesized by a set of spatial-statistical indexes representing a signature of the local surface morphology, and the relationships between the geomorphometric indexes and morphological features were evaluated. The studied basin was then segmented by means of a fuzzy k-means classification algorithm into different spatial subdomains characterized by specific local geomorphic signatures. This method allows us to objectively choose the optimal combination of the number of classes and degree of fuzziness characterizing the studied basin. Accordingly, the analysis can be adopted to the characteristics of the studied area, improving the interpretation of surface morphology and offering a valuable tool for comparative analysis between different geomorphological contexts. The results of the classification were compared with the actual surface morphology based on high-resolution orthophotos, aerial photos, and field surveys.

Validation of production data by using an AI-based classification methodology; a case in the Gulf of Mexico

Production data quality is a topic of general interest in the oil industry. It is often the only information available in sufficient amount in mature fields. With the development of information systems the oil companies have large volumes of data, but not all are reliable. The errors may have various origins, starting with problems in data acquisition systems. The existence of contaminated data may cause operational failures and lead to an inappropriate decision making process. A methodology for identification of contaminated data was applied in order to determine the quality of a production dataset of an oil field. To achieve this objective, a methodology based on data mining techniques, combining a fuzzy classification algorithm, neural network modeling and an iterative process, was applied to a real case, a database of an offshore field en México and the result was the classification of data: good, slightly contaminated or bad. The decline behavior of a well was evaluated, with good and slightly contaminated data and the results were appropriate. We concluded that this classification methodology based on intelligent algorithms generated a simple solution to the problem of quality determination of production data. We found that this methodology can be applied to any dataset. Of course, there is a degree of subjectivity in the methodology, and changing the restriction criteria for classification, the data quality determination may change. Furthermore, during the application of the proposed methodology, it was shown the effectiveness of data mining tools in the estimation of missing data.

Comparing object-based and pixel-based classifications for mapping savannas

The development of robust object-based classification methods suitable for medium to high resolution satellite imagery provides a valid alternative to ‘traditional’ pixel-based methods. This paper compares the results of an object-based classification to a supervised per-pixel classification for mapping land cover in the tropical north of the Northern Territory of Australia. The object-based approach involved segmentation of image data into objects at multiple scale levels. Objects were assigned classes using training objects and the Nearest Neighbour supervised and fuzzy classification algorithm. The supervised pixel-based classification involved the selection of training areas and a classification using the maximum likelihood classifier algorithm. Site-specific accuracy assessment using confusion matrices of both classifications were undertaken based on 256 reference sites. A comparison of the results shows a statistically significant higher overall accuracy of the object-based classification over the pixel-based classification. The incorporation of a digital elevation model (DEM) layer and associated class rules into the object-based classification produced slightly higher accuracies overall and for certain classes; however this was not statistically significant over the object-based using spectral information solely. The results indicate object-based analysis has good potential for extracting land cover information from satellite imagery captured over spatially heterogeneous land covers of tropical Australia.