KEYWORDS: Hyperspectral imaging, Data modeling, Principal component analysis, Sun, Single crystal X-ray diffraction, Code division multiplexing, Signal to noise ratio, Speckle, Autoregressive models, Feature extraction
We propose an unsupervised method for slight change extraction and detection in multitemporal hyperspectral image sequence. To exploit the spectral signatures in hyperspectral images, autoregressive integrated moving average and fitting models are employed to create a prediction of single-band and multiband time series. Minimum mean absolute error index is then applied to obtain the preliminary change information image (PCII), which contains slight change information. After that, feature vectors are created for each pixel in the PCII using block processing and locally linear embedding. The final change detection (CD) mask is obtained by clustering the extracted feature vectors into changed and unchanged classes using k-means clustering algorithm with k=2. Experimental results demonstrate that the proposed method extracts the slight change information efficiently in the hyperspectral image sequence and outperforms the state-of-the-art CD methods quantitatively and qualitatively.
In this paper, a novel super resolution (SR) method for remote sensing images based on compressive sensing (CS),
structure similarity and dictionary learning is proposed. The basic idea is to find a dictionary which can represent the
high resolution (HR) image patches in a sparse way. The extra information coming from the similar structures which
often exist in remote sensing images can be learned into the dictionary, so we can get the reconstructed HR image
through the dictionary in the CS frame due to the redundance in the image which has a sparse form in the dictionary. We
use K-SVD algorithm to find the dictionary and OMP method to reveal the sparse coding coefficient's location and value.
The difference between our method and the previous sample-based SR method is that we only use low-resolution image
and the interpolation image from itself rather than other HR images. Experiments on both optical and laser remote
sensing images show that our method is better than the original CS-based method in terms of not only the effect but also
the running time.
The recently-emerged compressive sensing (CS) theory goes against the Nyquist-Shannon (NS) sampling theory and
shows that signals can be recovered from far fewer samples than what the NS sampling theorem states. In this paper, to
solve the problems in image fusion step of the full-scene image mosaic for the multiple images acquired by a low-altitude
unmanned airship, a novel information mutual complement (IMC) model based on CS theory is proposed. IMC
model rests on a similar concept that was termed as the joint sparsity models (JSMs) in distributed compressive sensing
(DCS) theory, but the measurement matrix in our IMC model is rearranged in order for the multiple images to be
reconstructed as one combination. The experimental results of the BP and TSW-CS algorithm with our IMC model
certified the effectiveness and adaptability of this proposed approach, and demonstrated that it is possible to substantially
reduce the measurement rates of the signal ensemble with good performance in the compressive domain.
KEYWORDS: Data modeling, Data storage, 3D modeling, Laser development, Laser scanners, Clouds, Feature extraction, Reconstruction algorithms, Image segmentation, Data integration
Inventory checking is one of the most significant parts for grain reserves, and plays a very
important role on the macro-control of food and food security. Simple, fast and accurate method to obtain
internal structure information and further to estimate the volume of the grain storage is needed. Here in our
developed system, a special designed multi-site laser scanning system is used to acquire the range data clouds
of the internal structure of the grain storage. However, due to the seriously uneven distribution of the range
data, this data should firstly be preprocessed by an adaptive re-sampling method to reduce the data
redundancy as well as noise. Then the range data is segmented and useful features, such as plane and cylinder
information, are extracted. With these features a coarse registration between all of these single-site range data
is done, and then an Iterative Closest Point (ICP) algorithm is carried out to achieve fine registration. Taking
advantage of the structure of the grain storage being well defined and the types of them are limited, a fast
automatic registration method based on the priori model is proposed to register the multi-sites range data
more efficiently. Then after the integration of the multi-sites range data, the grain surface is finally
reconstructed by a delaunay based algorithm and the grain volume is estimated by a numerical integration
method. This proposed new method has been applied to two common types of grain storage, and experimental
results shown this method is more effective and accurate, and it can also avoids the cumulative effect of errors
when registering the overlapped area pair-wisely.
For multi-resolution land covering classification, many researches have focused on selecting and integrating appropriate
feature information from different spatial resolution data of the same area. However, when extending to large scale
problems, it is no surprise that low resolution data has worse performance, and high resolution data with wide coverage
area has more limitations. To solve this problem, a novel framework is presented which compounds multiple spatial
resolution data at arithmetic level without the limitation of full-scale multi-resolution data. The framework allows
integrating conditional random fields (CRFs) with "real" likelihood distribution. Discrete feature-likelihood mapping is
proposed to represent multi-to-single spatial correspondence. By considering spatial contextual information between
pixels, CRFs based classifier offers a robust and accurate framework. Our experiments show that the proposed method
can greatly improve the accuracy for large scale land covering classification applications.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.