Principal Component Images Research Articles

Synergistic use of ASTER, L-band ALOS PALSAR, and hyperspectral AVIRIS-NG data for exploration of lode type gold deposit – A study in Hutti Maski Schist Belt, India

In this study, we have explored the utility of advanced hyperspectral, dual-polarized microwave, and advanced multispectral data to delineate potential areas of mineralization associated with the lode type gold deposit in Hutti-Maski Schist Belt Karnataka, India. We derived different image-enhanced products using visible-near-infrared (VNIR) and short-wave infrared (SWIR) spectral bands of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) sensor to delineate major rock types of gold-bearing greenstone belt such as metabasalt, amphibolite, younger alkali feldspar rich granite, granodiorite, and tonalitic-granodioric gneiss (TTG). The principal component (PC) image composite of ASTER VNIR-SWIR bands was not only used to distinguish different granitoids but also confirmed their sharp instructive nature of contact with the greenstone rocks. We also used an index image composite using ASTER SWIR bands to discriminate granitoid, basalt and amphibolite. We could demarcate few lineaments extending through the mineralized zone (e.g., Hutti, Heera-Buddine mines, etc.) using the fused products of ASTER and two polarization channels of the Phased-array L band Synthetic Aperture Radar (PALSAR) sensor of the Advanced Land Observation Satellite (ALOS). All these image-enhanced products are validated using the reference geological map of the study area and field data collected along selected field -traverses. The mineralogical and petrographical data of the altered rocks from the study site revealed the formation of calcite, scapolite, and other minerals. Represenattive reflectance spectrum of the altered rock was used to spatially delineate the surface distribution of the patchy and small alteration zones using airborne hyperspectral data of the Advanced Visible Infrared Imaging Spectrometer Next Generation (AVIRIS-NG) sensor. In this regard, a unified spectral map of altered rocks was derived by integrating two alteration maps prepared independently using two different partial sub-pixel algorithms (constrained energy minimization and matched filtering). Occurrences of altered rocks in the vicinity to favorable geological lineaments and their spatial proximity to metavolcanic-younger felsic intrusive contacts suggest their importance as the potential sites for detailed exploration.

Eigen vector based analysis of Landsat OLI principal components and constrained energy minimization maps for discriminating iron enriched zones in banded iron formation (BIF) in Sidhi, Madhya Pradesh

In this study, we have used visible near infrared (VNIR) and shortwave infrared (SWIR) spectral bands of the Operational Land Imager (OLI) sensor of Landsat 8 satellite to delineate promising areas for iron exploration in the parts of Madhya Pradesh, India. OLI bands resampled laboratory reflectance spectra of Banded Iron Formation (BIF) were used as the reference to identify OLI spectral bands suitable in recording spectral features of Iron bearing minerals. Subsequently, BIFs were delineated in the principal component (PC) image prepared using those PC bands; which had opposite Eigen vector loading for the spectral bands; which recorded absorption minima and shoulder of the spectral feature of iron. Within the BIF, iron enriched zones were identified by integrating constrained energy minimization (CEM) maps of BIF, hematite and limonite-goethite. Integrated CEM map delineating iron enrichment zones was validated in the field based on measured Fe values of samples using X-ray Fluorescence Instrument.

Principal Component Analysis in Projection and Image Domains—Another Form of Spectral Imaging in Photon-Counting CT

We explore the feasibility of principal component analysis (PCA) as a form of spectral imaging in photon-counting CT. Using the data acquired by a prototype system and simulated by computer, we investigate the feasibility of spectral imaging in photon-counting CT via PCA for feature extraction and study the impacts made by data standardization and de-noising on its performance. The PCA in the projection domain maintains the data consistence that is essential for tomographic image reconstruction and performs virtually the same as that in the image domain. The first three primary components account for more than 99.99% covariance of the data. Within anticipation, the contrast-to-noise ratio (CNR) between the target and background in the first principal component image can be larger than that in the image generated from the data acquired in each energy bin. More importantly, the CNR in the first principal component image may be larger than that in the image formed by the summed data acquired in all energy bins (i.e., the conventional polychromatic CT image). In addition, de-noising can not only reduce the noise in images but also improve the effectiveness/efficiency of PCA in feature extraction. The PCA in either projection or image domain provides another form of spectral imaging in photon-counting CT that fits the essential requirements on spectral imaging in true color. The verification of PCA's feasibility in CT as a form spectral imaging and observation of its potential superiority in CNR over conventional polychromatic CT are meaningful in theory and practice.

Non-destructive determination of fat and moisture contents in Salmon (Salmo salar) fillets using near-infrared hyperspectral imaging coupled with spectral and textural features

The feasibility of using NIR hyperspectral imaging technique for predicting fat and moisture contents in salmon fillets was assessed by integrating both characteristic wavelengths and image texture features. Monte Carlo-uninformative variable elimination (MC-UVE) and successive projections algorithm (SPA) were combined to extract characteristic wavelengths. Ten textural features of the principal component images were obtained using histogram statistics (HS) and gray level co-occurrence matrices (GLCMs) methods. Three types of models (PLS, MLR and LS-SVM) were established based on different types of inputs including only characteristic wavelengths, only texture parameters and combination both characteristic wavelengths and textures, respectively. Compared among all models, LS-SVM model coupled with wavelength and texture information gave the highest prediction accuracies with RP = 0.9685, RMSEP = 1.1750, RPD = 4.0162 for fat and RP = 0.9688, RMSEP = 0.8021, RPD = 4.0357 for moisture, respectively. This study showed that the prediction accuracy can be improved by combining spectral features with textural features and the fusion of characteristic wavelength and textural features had better potential than single spectral information in assessing the fat and moisture contents of salmon. Satisfactory prediction results confirmed the suitability of NIR hyperspectral imaging for quantitative prediction of fat and moisture in salmon.

Combination of spectra and texture data of hyperspectral imaging for prediction and visualization of palmitic acid and oleic acid contents in lamb meat

The feasibility of combining spectral and textural information from hyperspectral imaging to improve the prediction of the C16:0 and C18:1 n9 contents for lamb was explored. 29 and 22 optimal wavelengths were selected for the C16:0 and C18:1 n9 contents, respectively, by conducting the variable combination population analysis–iteratively retaining informative variables (VCPA-IRIV) algorithm. To extract the textural features of images, a gray-level co-occurrence matrix (GLCM) analysis was implemented in the first principal component image. The least squares support vector machine (LSSVM) model and the partial least squares regression (PLSR) model were developed to predict the C16:0 and C18:1 n9 contents from the spectra and the fusion data. The distribution map was visualized using the best model with the imaging process. The results showed that the combination of the spectral and textural information of hyperspectral imaging coupled with the VCPA-IRIV algorithm had strong potential for the prediction and visualization of the C16:0 and C18:1 n9 contents of lamb.

Detection of common defects on mandarins by using visible and near infrared hyperspectral imaging

The presence of surface defects is one of the most influential factors in the quality and price of fresh fruit because consumers usually associate quality with a good appearance and without skin defects. Therefore, one of the main purposes of automatic detection of fruit quality is to differentiate between defective ones from sound fruits. However, the detection of defective fruits has always been a challenging task, especially the simultaneous detection of multiple types of defects. This work focuses on the development of multispectral image classification algorithm for detecting the common defects on mandarins based on the visible-near infrared (Vis-NIR) hyperspectral imaging technique. ‘Nanfeng’ mandarins with sound peel and four types of defects (i.e., anthracnose, scarring, decay and thrips scarring) were studied. Principal component analysis (PCA) was used to reduce hyperspectral data dimensionality with the goal of selecting several wavelengths that could be potentially used in an in-line multispectral imaging system. Two characteristic wavelength images at 680 nm and 715 nm in the visible spectral region were selected, and then the second principal component image (PC-2) and ratio image (Q680/715) based on these two characteristic wavelengths were used for defect detection and stem-end identification, respectively. Finally, the detection algorithm of defects was developed based on PC-2 image and ratio image (Q680/715) coupled with a simple thresholding method. For the investigated 356 independent test samples, classification accuracy of 96.63% indicated that the proposed multispectral image algorithm was effective for distinguishing between sound and defective ‘Nanfeng’ mandarins. Only two wavelength images were used in the algorithm, which was very helpful to develop a fast multispectral imaging system for on-line grading of mandarins.

Utilizing Landsat-8 and ASTER data in geologic mapping of hyper-arid mountainous region: case of Gabal Batoga area, South Eastern Desert of Egypt

Landsat-8 and ASTER images were utilized to derive geological information and mapping of a hyper-arid mountainous region. Principal component analyses and band rationing techniques in conjunction with petrography and field investigations enabled to produce a modified new geologic map of Gabal Batoga area, along the Red Sea coast of Egypt. The area is built by non-consanguineous diverse Pan-African basement rock units. These rock units are widely distributed in the Arabian-Nubian Shield. ASTER principal component images (PC2, PC5, PC7) and (PC2, PC3, PC8) as well as band ratio images (b4/b7, b3/b4, b2/b1) were effective in accurate lithological discrimination of the exposed rock units in the study area. The process was also enhanced and emphasized by Landsat-8 principal component images (PC2, PC4, and PC7) and band ratio (b6/b2, b6/b7, b6/b5 × b4/b5). The used methodology and techniques can be applicable in many similar arid and hyper-arid areas in the Arabian-Nubian Shield and elsewhere.

An integrated approach to modeling urban growth using modified built-up area extraction technique

Prediction of urban growth is often crucial in urban planning decisions. In this paper, we integrated the whole process of urban growth prediction by SLEUTH simulation for Dhaka Metropolitan Development Plan area of Bangladesh for the year of 2035. SLEUTH requires a rigorous preparation of five data inputs, i.e., slope, exclusion area, urban extent, road network and hillshade. This paper improvised the preparation of urban extent input using both Landsat-8 and Landsat-5 imagery. To increase the accuracy of urban area extraction from Landsat-8 images, we integrated normalized difference vegetation index and modified normalized difference water index with normalized difference built-up index. In the case of normalized difference built-up index, we used the principal component image of band 6 and 7 of Landsat-8 to include the effects of both bands. This technique to extract the built-up area increased the overall accuracy by 17.28% point. SLEUTH model ran through three calibration phases—coarse, fine and final—and an additional calibration was run to generate the forecasting coefficients. After the calibration phase, the best fit coefficient values were determined to run the prediction mode. The predicted outputs were derived as percentiles of development probability, from which a probability of above 90% was selected in this study. The prediction reveals that the urban extent of the study area is likely to increase by 158.66% from 2015 to 2035, and the designated conservation areas will significantly decrease during the same time period. This paper will provide researchers with an accurate and structured methodology to predict urban growth.

Spatial-Spectral Joint Classification of Hyperspectral Image With Locality and Edge Preserving

Hyperspectral image (HSI) classification is an important part of its processing and application. Aiming at the problems of high data dimensionality and high spatial neighborhood correlation in HSI classification, we propose a spatial-spectral joint classification method of HSI with locality and edge preserving in this article. First, the input HSI is normalized, and the feature is extracted by principal component analysis. The first principal component image is taken as the guidance image. Second, guided filtering is used to extract the spatial features of each band separately. Then, the extracted spatial features are superimposed, and low-dimensional embedding is completed through local Fisher discriminant analysis. Finally, the obtained low-dimensional embedded features are input into a random forest classifier to get classification results. The experimental results of two HSI show that the proposed method achieves higher classification accuracy than other related methods. In the case of randomly selecting 10% and 1% samples from each class of ground object as training samples, the overall classification accuracy is improved to 99.57% and 97.79%, respectively. This method effectively uses the spatial and local information of the image in low dimensional embedding, and preserves the boundaries of the ground objects, thus improving the classification effect.

Research on vegetation coverage optimisation of urban landscape based on vegetation index

In order to overcome the problem of fuzziness and randomness of remote sensing data, a vegetation coverage optimisation of urban landscape based on vegetation index is designed. The remote sensing image of vegetation cover is preprocessed by the principal component analysis. According to the fuzzy principal component image, GIS software is used as the platform of input data, and Fragstat software package is used to calculate the vegetation index of urban landscape. The NDVI value of shadow pixels is changed into background value, and Band Math is used on ENVI platform to get the NDVI data after removing the outliers. According to the principle of pixel dichotomy, the mixed pixel decomposition model is established to estimate the vegetation coverage and realise the vegetation coverage optimisation of urban landscape. The experimental results show that the vegetation coverage rate is 53%, the performance breakthrough is achieved and the pixel accuracy is more than 90%.

Identifying Differences Between a Straight Face and a Posed Smile Using the Homologous Modeling Technique and the Principal Component Analysis.

Recently, a homologous modeling method was developed to simulate 3D human body forms, which can visualize principal component analysis (PCA) results and facilitate its detailed comparison with results of previous method. Herein, we aimed to construct a homologous model of the face to identify differences between a straight face and a posed smile. Thirty-eight volunteers (19 males and 19 females, 38 straight faces and 38 posed smiles) with no medical history associated with a posed smile were enrolled. Three-dimensional images were constructed using the Homologous Body Modeling software and the HBM-Rugle; 9 landmarks were identified on the 3D-model surfaces. The template model automatically fitted into an individually scanned point cloud of the face by minimizing external and internal energy functions. Faces were analyzed using PCA; differences between straight faces and posed smiles were analyzed using paired t tests. Contribution of the most important principal component was 23.8%; 8 principal components explained >75% of the total variance. A significant difference between a straight face and a posed smile was observed in the second and the fourth principal components. The second principal component images revealed differences between a straight face and a posed smile and changes around the chin area with regard to length, shape, and anteroposterior position. Such changes were inclusive of individual differences. However, the fourth principal component image only revealed differences between a straight face and a posed smile; observed differences included simultaneous shortening of upper and lower eyelid length, evaluation of the nasal ala ase, swelling of the cheek area, and elevation of the mouth angle. Although these results were clinically apparent, we believe that this article is the first to statistically verify the same.Consequently, the homologous model technique and PCA are useful for evaluation of the facial soft-tissue changes.

Detection of bruised potatoes using hyperspectral imaging technique based on discrete wavelet transform

In the view of difficulties in bruise detection of potatoes, a texture recognition technique based on hyperspectral imaging and discrete wavelet transform was presented and discussed in this paper. Firstly, characteristic bands of intact and damaged potato samples based on the principal component analysis (PCA) were obtained. Then, the PCA method was used again to generate the principal component images that make it possible to distinguish the bruised potatoes from the healthy ones. In order to enhance the texture characteristics of the damaged samples, histogram equalization, mean filter and gradient method were utilized to preprocess the reference image. The paper achieved the best image damage recognition effects with the method of histogram equalization. Next, the preprocessed images were discomposed by the paper to obtain sub-band images using a two-level discrete wavelet transform (DWT), generating gray level co-occurrence matrix (GLCM) from sub-band images and texture feature data. Finally, the optimal parameters of gray level co-occurrence matrix were found out with the method of grid search and AdaBoost-Fisher algorithm to reduce the dimension of texture data generated under each parameter. The following results were obtained experimentally: the highest damage identification rate was 99.82%, the data dimensions were reduced from 56 to 4, the picture gotten from WT was reduced from 7 to 4, and the optimal parameter combination was obtained when the prediction effect was best after dimensionality reduction. The results show that the hyperspectral image texture recognition technology based on DWT-GLCM can accurately identify slightly damaged potato samples, providing a reference for potato on-line non-destructive testing.

Bilateral Kernel Extraction from PCA for Classification of Hyperspectral Images

The improved spatial and spectral resolution in the advanced Hyperspectral (HS) sensors results in images with rich information per pixel. Hence, the development of efficient spatial–spectral feature extraction (FE) techniques is very crucial for a proper characterization of the objects on ground. In this paper, an attempt has been made to develop a simple, yet effective spatial–spectral FE algorithm. In the proposed approach, the following steps are performed. First, Principal Component Analysis (PCA) was applied on the original hyperspectral image (HSI) and the most significant principal component was extracted. Then, the Bilateral Filter (BF), which acts as an edge-preserving filter, was applied on the selected principal component to extract kernel for each pixel in HSI. The extracted kernel bank is then applied on the original HSI. As in general, the principal component image is edge informative, and the BF is an edge-preserving filter; therefore, the extracted kernel bank can be applied on the original HSI to extract spatial–spectral features. Finally, with the help of these features, the performance of Support Vector Machine (SVM) classifier is evaluated. The proposed approach is validated on three popular hyperspectral data sets, namely, Indian Pines, Pavia University, and Botswana. The experimental results reveal that learning the edge information from a reference image (in the present context PCA) is quite essential, rather than applying the edge-preserving filters directly on the HSI. Theoretically, this holds true, as a unique edge (ground) information exists for an HSI, while in reality, the edges have variations due to variation in reflectance over bands.

Detection of early decayed oranges based on multispectral principal component image combining both bi-dimensional empirical mode decomposition and watershed segmentation method

Detection of early decay caused by fungal infections in citrus fruit still remains one of the major problems in the post-harvest processing and automatic quality grading. A new combination algorithm by merging multispectral principal component image, bi-dimensional empirical mode decomposition and image reconstruction as well as improved watershed segmentation was proposed to detect the early decay in oranges. Segmented principal component analysis based on three wavelength regions including visible and short wavelength near-infrared (500–1050 nm), visible (500–780 nm) and near-infrared (781–1050 nm) was performed to determine the optimal principal component (PC) image that was used to extract the effective wavelength images by weighting coefficient analysis. Seven wavelength images in the spectral region of 500–1050 nm were finally determined to build the multispectral PC images. The bi-dimensional empirical mode decomposition (BEMD) was used to remove noise in the multispectral PC images and further reconstruct images. An improved watershed segmentation method with morphological gradient reconstruction, marker extraction as well as image amendment, was proposed to segment decay regions in fruit by using the reconstructed multispectral PC images. All samples including 220 each of decayed and sound fruit were utilized to assess classification ability of the proposed combination algorithm. The results indicated that identification accuracies of decayed and sound fruit were 97.3% and 100%, respectively. The multispectral principal component image combining both bi-dimensional empirical mode decomposition and watershed segmentation method can be used as an effective tool for detection of early decayed oranges, and it was also promising for development of a fast and low-cost multispectral imaging system.

Evaluating salinity intrusion in estuaries using remote sensing data integrated in-situ observation

Saline intrusion reduces crop productivity, causes land degradation, decreases water quality, and severely affects agricultural production, the environment as well as livelihoods. Under the evolution of climate change and human activities from the upstream of the Mekong, the downstream areas of Dinh An and Cung Hau estuaries in Tra Vinh province are also significantly affected by saline intrusion from the East Sea. This article presents the integrated solution of remote sensing and GIS in monitoring and mapping salinity intrusion. The data used are Landsat 8 satellite images combined with salinity water monitoring data collected from actual observation stations during the dry season. Analysis showed that there was a statistically significant correlation between the observed salinity value of the water and the pixel value of the first principal component image. Simulation of spatial distribution from the study indicates that saline intrusion is now entering the interior with a distance from the estuary to about 30 - 48 km depending on the time of the dry season. The results of this study will assist managers in planning food safety strategies at the risk of saline intrusion.

Geospatial delineation and mapping of groundwater potential in Embu County, Kenya

Integration of Remote Sensing (RS) and the Geographical Information System (GIS) approaches in the field of groundwater resources management is a breakthrough. The RS and GIS geospatial approaches can enhance the assessment, monitoring, and conservation of groundwater resources. In this study, RS and GIS geospatial techniques were applied with the aim of identifying groundwater potential zones in Embu County, Kenya, based on selected multi influencing factors. Lineament layer was obtained by processing Landsat 8 ETM+ image using Principal Component Analysis in ENVI®4.7 and automatic extraction from Principal Component Image using the LINE module in Geomatica software. The resultant groundwater potential map showed that approximately 78% of the total area ranged from ‘high’ to ‘very high’ zones indicating that almost half of the study area has good groundwater potential. About 20% showed moderate potential while only 2% fell under the low potential zone. The proposed study approach can be used as a new way of modeling geospatial data for identification and mapping of groundwater potential zones. The study findings are useful to first-hand information planners and local authorities for assessment, planning, management and administration of groundwater resources in Embu County.

Prospecting for Clinoptilolite-Type Zeolite in a Volcano-Sedimentary Terrain Using ASTER Data: A Case Study from Alborz Mountains, Northern Iran

Zeolites are hydrated alumino-silicates of alkali metals and alkaline earth cations which occur in sedimentary and volcano-sedimentary terrains. In this study, visible–near-infrared and shortwave infrared data of ASTER were evaluated in prospecting for zeolite in part of the green tuff belt of the Alborz Mountains, northern Iran. The study area is dominantly covered by sedimentary and volcano-sedimentary rocks, in which zeolite minerals occur only in the Late Eocene vitric tuff. Principal components (PC) analysis and spectral information divergence (SID) were used to discriminate and map the sedimentary and volcano-sedimentary units and the zeolite-rich areas, respectively. The X-ray diffraction and reflectance spectroscopy results indicated that clinoptilolite is the major type of zeolite mineral in this area. Comparing a color composite image, produced from PC images 1–3–5 as R–G–B, with the published geological map and the field investigations indicated that major sedimentary and volcano-sedimentary units as well as their alluvial deposits were discriminated efficiently. Results of the SID method, using an image-derived spectrum of clinoptilolite as a reference, showed good agreements with the field observations. The results of this study indicated that ASTER data are useful for discriminating various sedimentary and volcano-sedimentary units as well as clinoptilolite-type zeolite-rich areas in arid and semiarid terrains.

Broadband reflectance, emittance spectroscopy and self-potential geophysical survey for targeting gold sulphide lode deposit in Bhukia, Rajasthan, India

In this study, we have analyzed the comparative potential of reflectance and emittance bands of advanced spaceborne thermal emission and reflection radiometer (ASTER) sensor for detecting surface exposures of Proterozoic rocks and gossan capping of gold mineralization. In this regard, few new image analysis products are derived. These are matched filtering factor-modified ratio (MFR) image and the inverse of a difference image of the ratio and matched filtering map (IDRM). These image products are used with the conventional ratio images, relative band depth images and index-based principal component (PC) images of reflectance and emittance bands of ASTER for delineating metasediments of the study area. We have found that reflectance band products are superior for mapping metasediments and gossan capping than the emissivity products, which are only suitable to enhance quartzite. Self-potential (SP) geophysical survey is carried out around the identified gossan site as SP is sensitive to sulphide mineralization. Results of the SP survey indicated presence of sulphide enrichment underneath the gossan.

Linearly uncorrelated principal component and deep convolutional image deblurring for natural images

Blind image deblurring of natural images still remains a demanding task. The traditional methods, pre-processes the uniform and non-uniform images with a deblurring algorithm and employs a low-rank prior algorithm. The rich textures do not possess enough similar patches in the deblurring process and this loss results in noisy images. Also, computational efficiency gets compromised during the performance of the succeeding process. In this study, the authors propose a novel method called, linearly uncorrelated principal component and deep convolution (LUPC-DC) for deblurring natural images. The natural images are first de-correlated with which good similar patches are extracted to generate a low-rank matrix by linearly uncorrelated principal component (PC) extraction. Then, the deep convolutional neural network model jointly extracts good similar patches and deblurs the first PCs. Eventually, good similar patches in the last PCs are suppressed using Hard Thresholding for computational efficiency. Analysis of concurrence performance of the algorithm confirms the viability of this method theoretically. In addition, simulation results and performance evaluations of image quality metrics are provided to assess the effectiveness of the proposed method. Moreover, the proposed method provides improvement in the peak-signal-to-noise ratio rate, success rate and reduction in the computation time for image deblurring.

Cognitive spectroscopy for wood species identification: near infrared hyperspectral imaging combined with convolutional neural networks.

From the viewpoint of combating illegal logging and examining wood properties, there is a contemporary demand for a wood species identification system. Several nondestructive automatic identification systems have been developed, but there is room for improvement to construct a highly reliable model. The present study proposes cognitive spectroscopy that combines near infrared hyperspectral imaging (NIR-HSI) with a deep convolutional neural network approach. We defined "cognitive spectroscopy" as a protocol that extracts features from complex spectroscopic data and presents the best results without human intervention. Overall, 120 samples representing 38 hardwood species were scanned using an NIR-HSI camera. A deep learning prediction model was built based on the principal component (PC) images obtained from the PC scores of hyperspectral images (wavelength range: 1000-2200 nm at approximately 6.2 nm interval). The results showed that the accuracy of wood species identification based on 6PC (PC1-PC6) images was 90.5%, which was considerably higher than the accuracy of 56.0% obtained with conventional visible images.