Showing papers on "Change detection published in 2002"

An adaptive semiparametric and context-based approach to unsupervised change detection in multitemporal remote-sensing images

Abstract: A novel automatic approach to the unsupervised identification of changes in multitemporal remote-sensing images is proposed. This approach, unlike classical ones, is based on the formulation of the unsupervised change-detection problem in terms of the Bayesian decision theory. In this context, an adaptive semiparametric technique for the unsupervised estimation of the statistical terms associated with the gray levels of changed and unchanged pixels in a difference image is presented. Such a technique exploits the effectivenesses of two theoretically well-founded estimation procedures: the reduced Parzen estimate (RPE) procedure and the expectation-maximization (EM) algorithm. Then, thanks to the resulting estimates and to a Markov random field (MRF) approach used to model the spatial-contextual information contained in the multitemporal images considered, a change detection map is generated. The adaptive semiparametric nature of the proposed technique allows its application to different kinds of remote-sensing images. Experimental results, obtained on two sets of multitemporal remote-sensing images acquired by two different sensors, confirm the validity of the proposed approach.

A unifying framework for detecting outliers and change points from non-stationary time series data

Abstract: We are concerned with the issues of outlier detection and change point detection from a data stream. In the area of data mining, there have been increased interest in these issues since the former is related to fraud detection, rare event discovery, etc., while the latter is related to event/trend by change detection, activity monitoring, etc. Specifically, it is important to consider the situation where the data source is non-stationary, since the nature of data source may change over time in real applications. Although in most previous work outlier detection and change point detection have not been related explicitly, this paper presents a unifying framework for dealing with both of them on the basis of the theory of on-line learning of non-stationary time series. In this framework a probabilistic model of the data source is incrementally learned using an on-line discounting learning algorithm, which can track the changing data source adaptively by forgetting the effect of past data gradually. Then the score for any given data is calculated to measure its deviation from the learned model, with a higher score indicating a high possibility of being an outlier. Further change points in a data stream are detected by applying this scoring method into a time series of moving averaged losses for prediction using the learned model. Specifically we develop an efficient algorithms for on-line discounting learning of auto-regression models from time series data, and demonstrate the validity of our framework through simulation and experimental applications to stock market data analysis.

Urban built-up land change detection with road density and spectral information from multi-temporal Landsat TM data

Abstract: In this article, Landsat TM images acquired during the same season from both 1984 and 1997 were analysed for urban built-up land change detection in Beijing, China, where great changes have taken place during the recent decades. To reduce the spectral confusion between urban 'built-up' and rural 'non built-up' land cover categories, we propose a new structural method based on road density combined with spectral bands for change detection. The road density represents one type of structural information while the multiple Landsat TM bands represent spectral information. Road density maps for both dates were produced using a gradient direction profile analysis (GDPA) algorithm and then integrated with spectral bands. Results from the spectral-structural postclassification comparison (SSPCC) and spectral-structural image differencing (SSID) methods were evaluated and compared with spectral-only change detection methods. The proposed SSPCC method greatly reduced spectral confusion and increased the accuracy of ...

An image change detection algorithm based on Markov random field models

Abstract: This paper addresses the problem of image change detection (ICD) based on Markov random field (MRF) models. MRF has long been recognized as an accurate model to describe a variety of image characteristics. Here, we use the MRF to model both noiseless images obtained from the actual scene and change images (CIs), the sites of which indicate changes between a pair of observed images. The optimum ICD algorithm under the maximum a posteriori (MAP) criterion is developed under this model. Examples are presented for illustration and performance evaluation.

Integrating intensity and texture differences for robust change detection

Abstract: We propose a novel technique for robust change detection based upon the integration of intensity and texture differences between two frames. A new accurate texture difference measure based on the relations between gradient vectors is proposed. The mathematical analysis shows that the measure is robust with respect to noise and illumination changes. Two ways to integrate the intensity and texture differences have been developed. The first combines the two measures adaptively according to the weightage of texture evidence, while the second does it optimally with additional constraint of smoothness. The parameters of the algorithm are selected automatically based on a statistic analysis. An algorithm is developed for fast implementation. The computational complexity analysis indicates that the proposed technique can run in real-time. The experiment results are evaluated both visually and quantitatively. They show that by exploiting both intensity and texture differences for change detection, one can obtain much better segmentation results than using the intensity or structure difference alone.

Impact of data integration technique on historical land-use/land-cover change: Comparing historical maps with remote sensing data in the Belgian Ardennes

Abstract: Historical reconstructions of land-use/cover change often require comparing maps derived from different sources. The objective of this study was to measure land-use/cover changes over the last 225 years at the scale of a Belgian landscape, Lierneux in Ardennes, on the basis of a heterogeneous time series of land cover data. The comparability between the land-cover maps was increased following a method of data integration by map generalisation. Two types of time series were built by integrating the maps either by reference to the initial map of the time series or by pair of successive maps. Land-cover change detection was performed on the initial time series without data integration and on the two types of integrated time series. Results reveal that land cover and landscape structure have been subject to profound changes in Lierneux since 1775, with an annual rate of change at the landscape level of up to 1.40%. The major land-cover change processes observed are expansion of grasslands-croplands and reforestation with coniferous species, leading to amore fragmented landscape structure. The annual rates of land-cover change estimated from integrated data are significantly different from the annual rates of change estimated without a prior integration of the data. There is a trade-off between going as far back in time as possibleversus performing change detection as accurately as possible.

Event detection in baseball video using superimposed caption recognition

Abstract: We have developed a novel system for baseball video event detection and summarization using superimposed caption text detection and recognition. The system detects different types of semantic level events in baseball video including scoring and last pitch of each batter. The system has two components: event detection and event boundary detection. Event detection is realized by change detection and recognition of game stat texts (such as text information showing in score box). Event boundary detection is achieved using our previously developed algorithm, which detects the pitch view as the event beginning and nonactive view as potential endings of the event. One unique contribution of the system is its capability to accurately detect the semantic level events by combining video text recognition with camera view recognition. Another unique feature is the real-time processing speed by taking advantage of compressed-domain approaches in part of the algorithms such as caption detection. To the best of our knowledge, this is the first system achieving accurate detection of multiple types of high-level semantic events in baseball videos.

Mining data streams under block evolution

Abstract: In this paper we survey recent work on incremental data mining model maintenance and change detection under block evolution. In block evolution, a dataset is updated periodically through insertions and deletions of blocks of records at a time. We describe two techniques: (1) We describe a generic algorithm for model maintenance that takes any traditional incremental data mining model maintenance algorithm and transforms it into an algorithm that allows restrictions on a temporal subset of the database. (2) We also describe a generic framework for change detection, that quantifies the difference between two datasets in terms of the data mining models they induce.

A comparison of land use and land cover change detection methods

Abstract: The principal objective of this project was to compare the results of different land use and land cover change detection approaches: traditional post-classification cross-tabulation, cross-correlation analysis, neural networks, knowledge-based expert systems, and image segmentation and object-oriented classification. A combination of both direct T1 to T2 change detection as well as post -classification analysis was employed. The test sites, located in the Stony Brook Millstone River Watershed in New Jersey, consisted of two 512 2 image blocks representative of the range of cover types and changes in the watershed. Nine la nd use and land cover classes were selected for analysis: Dense Urban, Residential, Turf & Grass, Agriculture, Deciduous Forest, Coniferous Forest, Water, Wetland, and Barren Land. Twenty-three possible change and no-change classes were identified. Landsat Thematic Mapper data from March 27, 1989 and September 3, 1989 represented conditions at T 1, and Landsat Enhanced Thematic Mapper data from May 4, 2000 and September 23, 1999 were used for T2. It was observed that there are merits to each of the four methods examined, and that, at this point of this research, no single approach can solve the land use change detection problem. This paper overviews the procedures used and presents some of the results of the change detection experiment.

Urban change detection based on an artificial neural network

Abstract: A method based on an artificial neural network (ANN) was developed to detect newly urbanized areas depicted in satellite sensor images. The method uses two Landsat Thematic Mapper (TM) images of a region acquired on different dates as input and supervises the ANN to classify the image data into 'from-to' classes. Principal component analysis (PCA) was applied to extract the salient features and to reduce the dimensionality of the input data prior to the ANN-based change detection. The Levenburg-Marquardt algorithm was used to accelerate the ANN's convergence. Experimental results from a case study show the ANN-based method requires only modest training time but can be 20-30% more accurate than post-classification comparison. PCA not only reduced the computational cost but improved the change detection accuracy as well. The results suggest the practical value of ANN-based change detection.

Unsupervised change-detection methods for remote-sensing images

Abstract: An unsupervised change detection problem can be viewed as a classification problem with only two classes corresponding to the change and no-change areas, respectively. Thanks to its simplicity, im- age differencing is a widely used approach to change detection. It is based on the idea of generating a difference image that represents the modulus of the spectral change vector associated with each pixel in the study area. To separate the ''change'' and ''no-change'' classes in the difference image, a simple thresholding-based procedure can be ap- plied. However, the selection of the best threshold value is not a trivial problem. We investigate and compare several simple thresholding meth- ods. The combination of the expectation-maximization algorithm with a thresholding method is also performed for the purpose of achieving a better estimate of the optimal threshold value. As an experimental inves- tigation, a study area damaged by a forest fire is considered. Two Land- sat TM images of the area acquired before and after the event are uti- lized to detect the burnt zones and to assess and compare the mentioned unsupervised change-detection methods. © 2002 Society of

Bayesian methods for change-point detection in long-range dependent processes

Abstract: We describe a Bayesian method for detecting structural changes in a long- range dependent process. In particular, we focus on changes in the long-range dependence parameter, d, and changes in the process level, l. Markov chain Monte Carlo (MCMC) methods are used to estimate the posterior probability and size of a change at time t, along with other model parameters. A time-dependent Kalman filter approach is used to evaluate the likelihood of the fractionally integrated ARMA model characterizing the long-range dependence. The method allows for multiple change points and can be extended to the long-memory stochastic volatility case. We apply the method to three examples, to investigate a change in persistence of the yearly Nile River minima, to investigate structural changes in the series of durations between intraday trades of IBM stock on the New York Stock Exchange, and to detect structural breaks in daily stock returns for the Coca Cola Company during the 1990s.

Anomaly detection in multidimensional data using negative selection algorithm

Abstract: While dealing with sensitive personnel data, the data have to be maintained to preserve integrity and usefulness. The mechanisms of the natural immune system are very promising in this area, it being an efficient anomaly or change detection system. This paper reports anomaly detection results with single and multidimensional data sets using the negative selection algorithm developed by Forrest et al. (1994).

Speaker change detection and tracking in real-time news broadcasting analysis

Abstract: This paper addresses the problem of real time speaker change detection and speaker tracking in broadcasted news video analysis. In such a case, both speaker identities and number of speakers are assumed unknown. A two-step speaker change detection algorithm, including potential change detection and refinement, is proposed. Speaker tracking is performed based on the results of speaker change detection. A Bayesian Fusion method is used to fuse multiple audio features to get a more reliable result. The algorithm has low complexity and runs in real-time with a very limited delay in analysis. Our experiments show that the algorithms produce very satisfactory results.

Motion-Based Video Representation for Scene Change Detection

Abstract: In this paper, we present a new framework to automatically group similar shots into one scene, where a scene is generally referred to as a group of shots taken place in the same site. Two major components in this framework are based on the motion characterization and background segmentation. The former component leads to an effective video representation scheme by adaptively selecting and forming keyframes. The later is considered novel in that background reconstruction is incorporated into the detection of scene change. These two components, combined with the color histogram intersection, establish our basic concept on assessing the similarity of scenes.

Classifying and Mapping General Coral-Reef Structure Using Ikonos Data

Abstract: Monitoring coral reef, seagrass, and sand features using contemporary remotely sensed data may prove to be a costeffective and time-efficient tool for reef surveys, change detection, and management. Previous attempts at subsuqface feature discrimination with satellite remote sensing have been limited in accuracy due to the effects of pixel mixing associated with poor spatial resolutions. While aerial reconnaissance may offer higher spatial resolutions than satellite sensors, it is often limited by the high costs of planning and implementing the missions, image rectification, area that can be covered, and repeat coverage. In this study, the Ikonos satellite with a 4- by 4-m spatial resolution in the multispectral bands was used as a tool for subsuqface feature identification. The Single-Image Normalization Using Histogram Adjustment was used for atmospheric corrections on the imagery. Classification was peqformed using bands 1, 2, and 3 (blue, green, and red) to maximize the water-penetration capabilities of the sensor. An accuracy assessment of the classification results was peqformed using in situ data collected at 62 points one day prior to the image being acquired. It was concluded that the Ikonos data were useful for discriminating sand, coral reef (at two depth intervals), and seagrass features (providing overall accuracies of 89 percent each for the two study areas). However, error still remained in discriminating small, diverse patch-reef features. This error (producers accuracy 67 percent) was found in the "Reef 5 5 m " class and was primarily attributed to the diversity of this spectral class, which may lead to a spectral signature based on the dominant cover type in a given pixel.

Adaptive random sampling for load change detection

Abstract: Timely detection of changes in traffic load is critical for initiating appropriate traffic engineering mechanisms. Accurate measurement of traffic is essential since the efficacy of change detection depends on the accuracy of traffic estimation. However, precise traffic measurement involves inspecting every packet traversing a link, resulting in significant overhead, particularly on high speed links. Sampling techniques for traffic load estimation are proposed as a way to limit the measurement overhead. In this paper, we address the problem of bounding sampling error within a pre-specified tolerance level and propose an adaptive random sampling technique that determines the minimum sampling probability adaptively according to traffic dynamics. Using real network traffic traces, we show that the proposed adaptive random sampling technique indeed produces the desired accuracy, while also yielding significant reduction in the amount of traffic samples. We also investigate the impact of sampling errors on the performance of load change detection.

Detection of abnormal change in a time series of graphs

Abstract: In the management of large enterprise communication networks, it becomes difficult to detect and identify causes of abnormal change in traffic distributions when the underlying logical topology is ...

Model-based sensor and actuator fault detection and isolation

Abstract: This work concerns the development of an analytical redundancy-based approach for detecting and isolating sensor, actuator, and component (i.e., plant) faults in complex dynamical systems, such as aircraft and spacecraft. The method is based on the use of constrained Kalman filters, which are able to detect and isolate such faults by exploiting functional relationships that exist among various subsets of available actuator input and sensor output data. A statistical change detection technique based on a modification of the standard generalized likelihood ratio statistic is used to detect faults in real time. The feasibility and efficacy of the approach is demonstrated through simulation in the context of a nonlinear jet engine control system.

Change detection for thematic mapping by means of airborne multitemporal polarimetric SAR imagery

Abstract: The paper addresses the detection of changes in multitemporal polarimetric radar images, focusing on small objects and narrow linear features. The images were acquired at C- and L-band by the airborne EMISAR system. It is found that the radar intensities are better suited for change detection than the correlation coefficient and the phase difference between the co-polarized channels. In the case of linear features, there is no obvious difference between the C- and L-bands , and slight variations of the flight tracks are acceptable at look angles larger than 35 degrees. Theoretical detection thresholds are evaluated from the statistical distribution of the intensity ratio due to speckle. For the linear features and for urban environments, the observed thresholds are larger than the theoretical predictions. This is interpreted as an effect of radar intensity variations on length scales smaller than the spatial image resolution. The signature of urban areas is very sensitive to deviations between the flight tracks, and the sensitivity is larger at C-band than at L-band. On the other hand, the intensity contrast between buildings and the urban background is smaller at L-band and larger at C-band. For change detection, thresholds may have to be chosen separately for each object class because the intensity ratios of different object classes vary differently as a function of time.

Online activity detection in a multiuser environment using the matrix CUSUM algorithm

Abstract: Early detection of the entrance of a new user into a multiuser communication channel is considered. The matrix CUSUM (cumulative-sum) test, a multialternative quickest change detection method, is introduced and then applied to this problem. The general behavior of this algorithm is described in analytical results and simulations.

Similarity measures for multisensor remote sensing images

Abstract: In this paper we will introduce several similarity measures between multisensor images. These measures are based on concepts such as statistical dependence or mutual information. The use of these measures allows for the design of image registration algorithms and automatic change detection techniques.

Detecting Abrupt Changes by Wavelet Methods

Abstract: The objective of this paper is to contribute to the methodology available for dealing with the detection and the estimation of the location of discontinuities in one-dimensional piecewise smooth regression functions observed in white Gaussian noise over an interval. Our approach is nonparametric in nature because the unknown function is not assumed to have any specific form. Our method relies upon a wavelet analysis of the observed signal and belongs to the class of "indirect" methods, where one detects and locates the change points prior to fitting the curve, and then uses ones favorite function estimation technique on each segment to recover the curve. We show that, provided discontinuities can be detected and located with sufficient accuracy, detection followed by wavelet smoothing enjoys optimal rates of convergence.

Scene change detection apparatus

Abstract: An apparatus for detecting a scene change is disclosed, which is provided for realizing functions such as a nonlinear browsing of a video, a video indexing and a key frame generation in a personal video recorder or video database. In the apparatus according to the present invention, accumulated histograms are extracted from the received two frames, and then a pixel value corresponding to a specific accumulated distribution of respective accumulated histograms is stored, thereby accurately detecting the scene change by comparing difference of pixel value lists. Also, an illumination change determining part for receiving first and second pixel lists from first and second pixel list extracting parts is additionally included so as to determine whether a brightness change of an image occurs due to a change of illumination conditions. Accordingly, it is possible to detect the scene change without any influence from changes of illumination, a camera flash or other optical elements.

Change detection thresholds for remotely sensed images

Abstract: The detection of change in remotely sensed images is often carried out by designating a threshold to distinguish between areas of change and areas of no change. The choice of threshold is often arbitrary however. The purpose of this paper is to offer a statistical framework for the selection of thresholds. The framework accounts for the facts that one carries out multiple tests of the null hypothesis of no change, when searching for regions of change over an image with a large number of pixels. Special attention is given to global spatial autocorrelation, which can affect the selection of appropriate threshold values.

Change detection and occlusion modes in road-traffic scenarios

Abstract: Change blindness phenomena are widely known in cognitive science, but their relation to driving is not quite clear. We report a study where subjects viewed colour video stills of natural traffic while eye movements were recorded. A change could occur randomly in three different occlusion modes—blinks, blanks and saccades—or during a fixation (as control condition). These changes could be either relevant or irrelevant with respect to the traffic safety. We used deletions as well as insertions of objects. All occlusion modes were equivalent concerning detection rate and reaction time, deviating from the control condition only. The detection of relevant changes was both more likely and faster than that of irrelevant ones, particularly for relevant insertions, which approached the base line level. Even in this case, it took about 180 ms longer to react to changes when they occurred during a saccade, blink or blank. In a second study, relevant insertions and the blank occlusion were used in a driving simulator environment. We found a surprising effect in the dynamic setting: an advantage in change detection rate and time with blanks compared to the control condition. Change detection was also good during blinks, but not in saccades. Possible explanation of these effects and their practical implications are discussed.