Md. Ali Hossain

Unsupervised feature extraction based on a mutual information measure for hyperspectral image classification

Finding the most informative features from high dimensional space for reliable class data modeling is one of the most challenging problems in hyperspectral image classification. The problem can be address using two basic techniques: feature selection and feature extraction. One of the most popular feature extraction methods is Principal Component Analysis (PCA), however its components are not always suitable for classification. In this paper, we present a feature reduction method (MI-PCA) which uses a nonparametric mutual information (MI) measure on the components obtained via PCA. Supervised classification results using a hyperspectral data set confirm that the new MI-PCA technique provides better classification accuracy by selecting more relevant features than when using either PCA or MI on the original data.

One-Class Oriented Feature Selection and Classification of Heterogeneous Remote Sensing Images

Information extraction from spatial big data faces challenges in data relevancy analysis and heterogeneous data modeling. When the interested targets are more than one, the relevant analysis is often compromised. In this paper, a one-class oriented approach for effective feature selection and classification of remote sensing images is proposed. Mutual information (MI) is used as the feature selection criterion to cope with a wide range of data types. Then a cluster space (CS) representation is applied to model multimodal data and classifies each target class in turn. Hyperspectral and LiDAR data sets were used in the experiments. The test results demonstrate the advantage in terms of classification accuracies by focusing on one class at a time as compared to considering all classes simultaneously in classification.

Reconstruction of satellite images by multi-temporal gradient based sequential prediction

The presence of atmosphere can cause obstructions to satellite remote sensing by absorbing and scattering the electromagnetic energy. Therefore, transmittance of the atmosphere is an important factor to consider in a sensing system design1. Also the weather conditions such as the levels of the haze, dust or mist present in the environment, introduce distortion. Relative distributions of the brightness values of images can be different depending on the seasonal effect, termed radiometric inconsistency, which is solely dependent upon the solar radiation, illumination and reflectivity effects of the object and the conditions of the atmosphere during that time. Since they change frequently, multi-temporal data have low consistency over time. The inconsistency present in the remote sensed satellite images taken for sequential analysis can cause misguiding informaiton widely used in a range of oceanographic, terrestrial and atmospheric applications, such as land-cover mapping, environmental monitoring and disaster management. Degraded multi-temporal images needs to be checked and reconstructed before it can be used. In this paper a gradient adjusted temporal prediction approach has been used to predict or approximate the recent corrupted image using previous reference image.

Histogram based water quality assessment in satellite images

Water is one of the most precious resources of our environment. This water body is often faced the quality questions because of being polluted by ammonia, chemical wastes, sulfur dioxide from power plants, fertilizers containing nutrients-nitrates and phosphates, sediment, phytoplankton, etc. So, it is very necessary to assess quality of different water bodies. In this study, satellite images have been used for water quality measurement using histogram comparisons. A satellite image has been chosen to use as the original image whose water body has been considered as a standard of clear water body. Then this clear water body has been separated from other features using clustering at first to be used as standard ones and their number of pixels in percentage has been counted. Those images which contain the same percent of water bodies as the perspective standard ones can be verified by comparing their histograms. Euclidean distance has been measured between the standard and tested images histograms. A tolerance level has been taken to assess the water quality as excellent, better, good, bad and poor. Finally if the distance falls within the tolerance level the water body can be categorized as excellent, better, good, bad or poor based on their degree of purity.

Closest class measure based subspace detection for hyperspectral image classification

The objective of this study is to develop a hybrid nonlinear subspace detection technique in which Kernel Principal Component Analysis (KPCA) is combined with a Closest Class Pair (CCP) measure for the task of hyperspectral image classification. In the proposed approach, KPCA is applied first to generate the new features from original dataset then the CCP is applied to rank the features that are able to separate the complex or overlapping classes. Finally, the two ranked scores such as KPCA and CCP are combined to select a subset of features which is relevant and able to provide better discrimination among the input classes of interest. Experiments are performed on a real hyperspectral image acquired by the NASA Airborne Visible Infrared Imaging Spectrometer (AVIRIS) sensor and it can be seen that the proposed approach obtained the best classification accuracy 84.58%.

Effective subspace detection based on cross cumulative residual entropy for hyperspectral image classification

Remote sensing hyperspectral images are blessings of technology through which the ground objects can be detected effectively with the cost of computer processing. For classification of hyperspectral images finding an effective subspace is very important to classify them efficiently. In recent years, many researchers have drawn their interest to extract data more effectively from hyperspectral dataset. In this research, an approach has been proposed to find the effective subspace by measuring the relevance of individual features through Cross Cumulative Residual Entropy from the Principal Component images. The Support Vector Machine has been used as the classifier for the assessment of the feature reduction performance. Experiment has been completed on real hyperspectral dataset and achieved 97% of accuracy which is better than the standard approaches studied.

Satellite image compression using integer wavelet regression

Multi-temporal Image Compression is now an immerging field considering the fact that terabytes of data is now available for download every day. Evantualy temporal data compression is becoming a critical issue for fast data transmission. Many works have been done regarding compression in the field of satellite images that utilizes the spectral and spatial redundancies using predictive and transformed based procedures for lossless data compression, but, most of the contributions are on individual data or on single data. The main objective of this paper is to exploit the temporal correlation between the images. The recent image will be predicted from the historical image that is already available to the user. This will substantially reduce the load in transmitting the images. This paper actually emphasis on the process of increasing temporal correlation, which consequently improves the compression gain. In sequential transmission, the transmitted data will be used in future as a reference. Therefore, a new lossless approach has been introduced where reversible integer wavelet transformation is used to improve the temporal correlation. The experimented results show that the proposed method outperformed many state of art lossless approaches including JPEG2000.

Feature mining for effective subspace detection and classification of hyperspectral images

Hyperspectral image analysis is becoming an important field of research interest because of its wide range of applications in ground surface identification. New technology is being developed to capture hyperspectral images to cover more spectral bands and finer spectral resolution but also increases challenges to process those images for high correlation among data and both the spectral and spatial redundancy. This paper proposed a feature mining approach for the relevant feature selection as well as efficient classification of the hyperspectral dataset. Principal Component analysis and Mutual Information is two widely used feature reduction techniques utilized in conjunction for the feature reduction of the remote sensing data set. The kernel support vector machine classifier is used to assess the effectiveness of the detected subspace for classification. The proposed feature mining approach is able to achieve 99.3% classification accuracy on real hyperspectral data which higher than the standard approaches studied.

Statistical Similarity Based Change Detection for Multitemporal Remote Sensing Images

Change detection (CD) of any surface using multitemporal remote sensing images is an important research topic since up-to-date information about earth surface is of great value. Abrupt changes are occurring in different earth surfaces due to natural disasters or man-made activities which cause damage to that place. Therefore, it is necessary to observe the changes for taking necessary steps to recover the subsequent damage. This paper is concerned with this issue and analyzes statistical similarity measure to perform CD using remote sensing images of the same scene taken at two different dates. A variation of normalized mutual information (NMI) as a similarity measure has been developed here using sliding window of different sizes. In sliding window approach, pixels’ local neighborhood plays a significant role in computing the similarity compared to the whole image. Thus the insignificant global characteristics containing noise and sparse samples can be avoided when evaluating the probability density function. Therefore, NMI with different window sizes is proposed here to identify changes using multitemporal data. Experiments have been carried out using two separate multitemporal remote sensing images captured one year apart and one month apart, respectively. Experimental analysis reveals that the proposed technique can detect up to 97.71% of changes which outperforms the traditional approaches.

Weighted normalized mutual information based change detection in remote sensing images

Change detection from remote sensing images is getting more interest now a days because of abrupt changes in earth surface due to natural disasters or man-made activities. So its an important research question of how to extract relevant information about the changes due to rainfall, droughts, flooding, destroying land cover areas and so on. This problem has been studied in some research however many of these did not consider the nonlinear relationship while detecting the changes. In this research, above limitation has been addressed and Weighted Normalized Mutual Information (WNMI) is utilized for the improvement. The WNMI technique has been applied between the reference and target images to find out the changes. Thus the changes between every object of the given dataset have been identified and able to observe the damage of any specific area as well as its subsequent recovery. Weighting has been done to count significance at the pixel level. The proposed technique can detect the changes more effectively than the traditional mutual information approach. Experimental analysis is carried on real remote sensing images and it is found that the proposed method can detect more than 96% of changes which is much better than the standard benchmark techniques.