Hilal Soydan

A local extrema based method on 2D brightness temperature maps for detection of archaeological artifacts

Archaeological studies using computer vision based analysis methods on thermal imageries mainly lack an important stage of pointwise detection of artifact positions, which is needed for the automation of the system in a generic application. In this paper, we propose a pointwise detection method working in the thermal range of hyperspectral band for archaeological artifacts. The proposed method first optimally converts a given 3D hyperspectral image of the searched scene into a 2D brightness-temperature map by minimizing the mean square error (MSE) between the spectral radiance of a pixel and the Planck curves generated at different temperatures. The local maxima and minima are then found on the resulting 2D map as the candidate points. Finally, a score assignment is performed on the candidate points by using their temperature difference with respect to their neighborhood. The results on the thermal images taken from a test scene have indicated a good correlation between the extrema points and artifact positions.

Signature based vegetation detection on hyperspectral images

In this study, the contribution of utilizing hyperspectral unmixing algorithms on signature based target detection algorithms is studied. Spectral Angle Mapper (SAM), Spectral Matched Filter (SMF) and Adaptive Cosine Estimator (ACE) algorithms are selected as target detection methods and the performance change related to the target spectral acquisition is evaluated. The spectral signature of the desired target, corn, is acquired from ASD hyperspectral library as well as from the hypespectral unmixing endmembers with a minimum angular distance to ASD signature. It is seen that the performance of the corn detection has increased significantly with the utilization of the closest endmember extracted from the hyperspectral data cube. Among all methods, SAM has been designated as the most successful method based on the Receiver Operating Characteristics (ROC) curves.

Hydrocarbon microseepage mapping using signature based target detection techniques

In this paper, we compare the conventional methods in hydrocarbon seepage anomalies with the signature based detection algorithms. The Crosta technique [1] is selected as a basement in the experimental comparisons for the conventional approach. The Crosta technique utilizes the characteristic bands of the searched target for principal component transformation in order to determine the components characterizing the target in interest. Desired Target Detection and Classification Algorithm (DTDCA), Spectral Matched Filter (SMF), and Normalized Correlation (NC) are employed for signature based target detection. Signature based target detection algorithms are applied to the whole spectrum benefiting from the information stored in all spectral bands. The selected methods are applied to a multispectral Advanced SpaceBorne Thermal Emission and Radiometer (ASTER) image of the study region, with an atmospheric correction prior to the realization of the algorithms. ASTER provides multispectral bands covering visible, short wave, and thermal infrared region, which serves as a useful tool for the interpretation of the areas with hydrocarbon anomalies. The exploration area is selected as Gemrik Anticline which is located in South East Anatolia, Adıyaman, Bozova Oil Field, where microseeps can be observed with almost no vegetation cover. The spectral signatures collected with Analytical Spectral Devices Inc. (ASD) spectrometer from the reference valley [2] have been utilized as an input to the signature based detection algorithms. The experiments have indicated that DTDCA and MF outperforms the Crosta technique by locating the microseepage patterns along the mitigation pathways with a better contrast. On the other hand, NC has not been able to map the searched target with a visible distinction. It is concluded that the signature based algorithms can be more effective than the conventional methods for the detection of microseepage induced anomalies.

Identification of hydrocarbon microseepage induced alterations with spectral target detection and unmixing algorithms

Abstract Hydrocarbon micro and macro seeps alter chemical and mineral composition of the Earth’s surface, providing prospects for detection with remote sensing tools. There have been several studies focusing on mapping these anomalies by utilizing ever evolving multispectral and hyperspectral imaging instruments, which has proven their capacity for mapping both hydrocarbons and hydrocarbon-induced alterations so far. These studies broadly comprise of methods like calculating band ratios, spectral angle mapping, spectral feature fitting, and principal component analysis as detection techniques. However, there is a lack of concentration on advanced signature based detection algorithms and unmixing methods for mapping surface manifestations of hydrocarbon microseeps. Signature based detection algorithms utilize target spectra to correlate with each pixel’s spectrum in order to allocate possible target locations. Unmixing methods, on the other hand, require no input spectra beforehand, aiming to resolve each pixel’s spectral constituents and their corresponding abundance fractions. In this paper, the potential of all these methods in mapping microseepage related anomalies are evaluated by implementing and comparing them for Gemrik Anticline, one of the prospective hydrocarbon exploration fields in Turkey. Hence, it provides a complete knowledge on determination surface manifestations of hydrocarbon microseeps with the help of well known supervised target detection algorithms and hyperspectral unmixing algorithms. The study area is located in the Southeastern Anatolia, between the cities of Adiyaman and Şanliurfa. The spectral signatures were collected with Analytical Spectral Devices Inc. (ASD) spectrometer during the field studies conducted by Avcioglu (2010), to be utilized as an input to the signature based detection algorithms as well as a reference to select the related abundance map among the outputs of unmixing methods. Advanced Space Borne Thermal Emission and Radiometer (ASTER) image of the study region, with an atmospheric correction before running the algorithms, is selected for the applications. Among the applied algorithms, Simplex Identification via Split Augmented Lagrangian (SISAL) is selected as a base of comparison, as it possess minimum calculated error metrics in the experiments. Another unmixing method, the Minimum Volume Simplex Algorithm (MVSA), and signature-based techniques, Desired Target Detection and Classification Algorithm (DTDCA) & Spectral Matched Filter (SMF) follow the success of the SISAL, respectively. The Crosta technique, which is performed as a conventional approach for experimental comparisons, has also shown its capability, succeeding these algorithms. The study provides an overall assessment for methodologies to be used for hydrocarbon microseepage mapping, which also serves guidance for further exploration studies in the region. The potential of ASTER data for hydrocarbon-induced alterations is also emphasized as a cost effective tool for the future applications.

Spatio-temporal anomaly detection for environmental impact assessment: a case of an abandoned coal mine site in Turkey

The main purpose of this research is to determine the anomalies regarding with the coal mining operations in an abandoned coal mine site in central Anatolia by multi-temporal image analysis of Landsat 4-5 surface reflectance data. A well-known anomaly detection algorithm, Reed-Xioli (RX), which calculates square of Mahalanobis metrics to calculate the likelihood ratios by normalizing the difference between the test pixel and the background to allocate anomaly pixels, is implemented across the time series. The experimental results reveal especially the profound land use – land cover change in time series, pointing out critically abandoned regions that need immediate rehabilitation action. The rate of anomaly scores together with their relation to mine development over the focused time spectrum discloses a linearity trend as of the operations are ceased at the end of 1990s, which is indicative of the capacity of the applied method. The performance of the algorithm is also quantified with Receiver Operating Characteristics (ROC) curves and precisionrecall graphs to quantify its capability on Landsat Thematic Mapper (TM) multispectral image series. The resulting plots show the increasing capability of the hyperspectral anomaly detection technique in multi-temporal data set, with a steady and slight increase in performance between 2000 and 2012 after the end of the mining activities, which substantiates the success of global RX algorithm to identify the mining-induced land use and land cover anomalies.

Total carbon mapping with hyperspectral unmixing techniques

Depending on the ground sampling distance of a remote sensor, a pixel of a spectral data cube is represented as a combination of the reflected signals of the materials which constitutes the observed pixel. Hyperspectral unmixing algorithms model the pixel of a data cube to determine and extract the spectral signatures of its components, namely endmembers, with their corresponding abundance fractions. This study first reviews the interaction and mitigation mechanisms of heavy metals with carbon content in soil, specifically due to coal mining activities and thermal plants. Such mechanism is then investigated with hyperspectral unmixing techniques by producing total carbon maps for an abandoned coal mine site. The utilized data for the study area is obtained on August 2013 with multispectral Worldview-2 satellite sensor. The acquired image is orthorectified and atmospherically corrected for radiance to reflectance conversion prior to the analysis. The soil samples are mainly collected from the problematic regions in terms of soil pollution. The samples are analyzed with LECO TrueSpec CHN_S device to measure total carbon levels, which are employed as ground truth to assess the performance of unmixing algorithms. The resulting abundance maps for carbon content are found to have a high compatibility with each other and the ground truth data, which effectively point out the regions of high carbon content.

Hyperspectral unmixing based vegetation detection with segmentation

This paper presents a vegetation detection application with semi-supervised target detection using hyperspectral unmixing and segmentation algorithms. The method firstly compares the known target spectral signature from a generic source such as a spectral library with each pixel of hyperspectral data cube employing Spectral Angle Mapper (SAM) algorithm. The pixel(s) with the best match are assumed to be the most likely target vegetation locations. The regions around these potential target locations are further analyzed via hyperspectral unmixing techniques to obtain the real spectra in the image. The abundance fractions are evaluated so as to compare the algorithm performance with those of other methods. As a post processing technique meanshift segmentation algorithm utilized.

A comparative study of hyperspectral anomaly and signature based target detection methods for oil spills

Hyperspectral target detection methods have until now progressed mainly on two paths in remote sensing research. The first approach, anomaly detection methods, use the difference of a local region with respect to its neighborhood to analyze the image without using any prior information of the searched target. The second approach on the other hand uses a previously obtained signature of the target, which uniquely represents the targets characteristics with respect to the spectral wavelengths. The signature of the target is matched with the pixels of the acquired image to decide on the existence and location of the searched target. These two approaches provide crucial information to detect oil spills to monitor environmental pollution. In this paper, we aim to use and compare anomaly and signature based target detection approaches for the identification of oil slicks. The study area is selected as the Gulf of Mexico, where one of the worst marine oil spill accidents in the history of the petroleum industry occurred in April 2010. The results indicate that signature based algorithms have a better performance in detecting, locating, and quantifying oil spills compared to the anomaly detection methods. Among the anomaly detection methods, the Gaussian Kernel Reed-Xiaoli (RX) method shows also a close performance to signature based methods, although it requires very long execution times on the down side.

Oil spill determination with hyperspectral imagery: A comparative study

Hyperspectral target detection methods have until now progressed mainly on two paths in remote sensing research. The first approach, anomaly detection methods, use the difference of a local region with respect to its neighborhood to analyze the image without using any prior information of the searched target. The second approach on the other hand uses a previously obtained signature of the target, which uniquely represents the targets reflection characteristics with respect to the spectral wavelengths. The signature of the target is matched with the pixels of the acquired image to decide on the existence and location of the searched target. These two approaches provide crucial information to detect oil spills to monitor environmental pollution. In this paper, we aim to use and compare anomaly and signature based target detection approaches for the identification of oil slicks. The study area is selected as the Gulf of Mexico, where one of the worst marine oil spill accidents in the history of the petroleum industry occurred in April 2010. The results indicate that signature based algorithms have a better performance in detecting, locating, and quantifying oil spills compared to the anomaly detection methods. Among the anomaly detection methods, the Gaussian Kernel Reed-Xiaoli (RX) method shows also a close performance to signature based methods, although it requires very long execution times on the down side.

Analysis of land use land cover changes for an abandoned mine site

The aim of this study is to the evaluate land use and the land cover changes of an abandoned coal mine in Central Anatolia. The mining activity in the region was started at 1987 and after working for 18 years, all the rights of the mine was passed to an another company which after a while the came up with no coal production since its concession was cancelled on February 2008. Unfortunately, during the life of the mine, there were no mine closure and reclamation activities on the field. There was only a limited afforestation work on the dump site performed by provincial special administration. As there were no appropriate closure and reclamation implementations during and after the mining activity, major environmental impacts detected on the mine field. This study investigates the land use land cover change due to mining operations between 1987 and 2000. Landsat ETM+ imagery is selected. Maximum Likelihood estimation based on the Bayesian theorem is used for classification purposes. Change maps are produced to observe the differences in land use and land cover changes in the study area in comparison with the baseline of the region.